Downloads · 30 days
22
24% of all-time downloads
Qwopus3.6-27B-v2-oQ6-fp16-mtp
RoelV/Qwopus3.6-27B-v2-oQ6-fp16-mtp
Qwopus3.6-27B-v2-oQ6-fp16-mtp is a image-text-to-text model from RoelV. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; border: 1px solid cbd5e1; border-radius: 16px; box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.05), 0 4px 6px -2px rgba(0, 0, 0,…
All-time downloads
91
Public
Parameters
6.6B
23.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors23.7 GB · 100%
Parameter types
How the weights are stored.
U325.2B · 79%
Try a prompt
Base models
Model card
From the Hugging Face model README
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; border: 1px solid #cbd5e1; border-radius: 16px; box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.05), 0 4px 6px -2px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff; margin-bottom: 30px;">
<div style="background: linear-gradient(135deg, #7c3aed 0%, #4c1d95 100%); padding: 24px; color: white;">
<div style="display: flex; align-items: center; justify-content: space-between; flex-wrap: wrap; gap: 10px;">
<h1 style="margin: 0; font-size: 26px; font-weight: 800; display: flex; align-items: center; gap: 12px; color: white; border: none;">🪐 Qwopus3.6-27B-v2</h1>
<span style="background: #10b981; color: white; font-size: 11px; font-weight: 700; padding: 4px 10px; border-radius: 20px; text-transform: uppercase; letter-spacing: 0.5px;">SFT Release</span>
</div>
<p style="margin: 8px 0 0 0; font-size: 14px; color: #ddd6fe; font-weight: 500;">Reasoning-Enhanced Dense Language Model Fine-Tuned on Qwen3.6-27B</p>
</div>
<div style="display: flex; gap: 8px; flex-wrap: wrap; padding: 12px 24px; background: #f8fafc; border-bottom: 1px solid #e2e8f0;">
<span style="background: #f3e8ff; color: #6b21a8; font-size: 11px; font-weight: 700; padding: 4px 10px; border-radius: 20px; border: 1px solid #e9d5ff;">🧬 Trace Inversion & Negentropy</span>
<span style="background: #dbeafe; color: #1e40af; font-size: 11px; font-weight: 700; padding: 4px 10px; border-radius: 20px; border: 1px solid #bfdbfe;">🧠 27B Parameters</span>
<span style="background: #e0f2fe; color: #0369a1; font-size: 11px; font-weight: 700; padding: 4px 10px; border-radius: 20px; border: 1px solid #bae6fd;">🔥 3-Stage Curriculum SFT</span>
<span style="background: #d1fae5; color: #065f46; font-size: 11px; font-weight: 700; padding: 4px 10px; border-radius: 20px; border: 1px solid #a7f3d0;">🛠️ Vision & Tool-use Support</span>
</div>
<div style="padding: 24px; display: flex; flex-direction: column; gap: 20px;">
<div style="background: #f5f3ff; border-left: 5px solid #7c3aed; padding: 16px; border-radius: 0 8px 8px 0;">
<h3 style="margin: 0 0 8px 0; font-size: 15px; color: #6d28d9; font-weight: 700; display: flex; align-items: center; gap: 6px;"><span>💡</span> What is Qwopus3.6-27B-v2?</h3>
<p style="margin: 0; font-size: 13px; color: #334155; line-height: 1.6;">🪐 <b>Qwopus3.6-27B-v2</b> is a reasoning-enhanced dense language model built on top of <b>Qwen3.6-27B</b>. By leveraging a multi-stage curriculum learning pipeline and augmented with <b>Trace Inversion</b> datasets (claude-opus-4.6/4.7-traceInversion), it reverse-engineers the compressed "Reasoning Bubbles" of commercial LLMs into structured, step-by-step synthetic reasoning traces, successfully eliminating logical shortcuts and knowledge fractures.</p>
</div>
<div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 15px; margin-top: 10px;">
<div style="border: 1px solid #e2e8f0; padding: 14px; border-radius: 8px; background: #fafafa; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);">
<span style="font-weight: 700; color: #6b21a8; font-size: 12px; display: block; margin-bottom: 6px; text-transform: uppercase; letter-spacing: 0.5px;">🧩 Structured Reasoning</span>
<span style="font-size: 13px; color: #4b5563; line-height: 1.5;">Injects reconstructed deep CoT chains to eliminate logical shortcuts via Trace Inversion.</span>
</div>
<div style="border: 1px solid #e2e8f0; padding: 14px; border-radius: 8px; background: #fafafa; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);">
<span style="font-weight: 700; color: #6b21a8; font-size: 12px; display: block; margin-bottom: 6px; text-transform: uppercase; letter-spacing: 0.5px;">🪶 Style Consistency</span>
<span style="font-size: 13px; color: #4b5563; line-height: 1.5;">Enforces strict constraints on the format and convergence of <code><think></code> tags.</span>
</div>
<div style="border: 1px solid #e2e8f0; padding: 14px; border-radius: 8px; background: #fafafa; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);">
<span style="font-weight: 700; color: #6b21a8; font-size: 12px; display: block; margin-bottom: 6px; text-transform: uppercase; letter-spacing: 0.5px;">🔁 Distillation Alignment</span>
<span style="font-size: 13px; color: #4b5563; line-height: 1.5;">Ensures high-quality cross-source SFT data alignment to narrow the capacity gap.</span>
</div>
<div style="border: 1px solid #e2e8f0; padding: 14px; border-radius: 8px; background: #fafafa; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);">
<span style="font-weight: 700; color: #6b21a8; font-size: 12px; display: block; margin-bottom: 6px; text-transform: uppercase; letter-spacing: 0.5px;">⚡ RL Scalability</span>
<span style="font-size: 13px; color: #4b5563; line-height: 1.5;">Sets up a stable formatting pipeline optimized for downstream Reinforcement Learning (RL).</span>
</div>
</div>
</div>
</div>
💡 1. Base Model, Training Library & Cooperation
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; display: flex; flex-direction: column; gap: 20px; margin-bottom: 30px;"> <!-- Card 1: Base Model Overview --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #7c3aed 0%, #5b21b6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>🧠</span> 1.1 Base Model Specifications (Qwen3.6-27B) </div> <div style="padding: 16px;"> <p style="margin: 0 0 16px 0; font-size: 13px; color: #334155; line-height: 1.6;"> <b>Qwen3.6-27B</b> is a state-of-the-art dense large language model developed by Alibaba Cloud. Boasting 27 billion parameters, this base model natively supports long-context modeling and is engineered for agentic workflows, complex logical reasoning, and multimodal capabilities. </p> <table style="width: 100%; border-collapse: collapse; font-family: inherit; font-size: 13px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-weight: bold; width: 30%;">Attribute</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left;">Specifications & Details</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">🧠 Architecture</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Dense Transformer / 27 Billion Parameters</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">🏢 Developer</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Alibaba Cloud (DAMO Academy)</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">📄 Context Window</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Native Support Up to 32K / 128K context length</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px;">🎯 Focus Domains</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Agentic Coding, Deep Logic Reasoning, Multimodal Tasks (Vision & Tool-use)</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px;">🧬 Distillation Strategy</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Cross-source SFT alignment and multi-teacher distillation to close the capability gap.</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px;">⚡ RL Scalability</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">Optimized for downstream Reinforcement Learning alignment and self-critical learning loops.</td> </tr> </tbody> </table> </div> </div> <!-- Card 2: Hardware Cooperation --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #10b981 0%, #047857 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>🧪</span> 1.2 Hardware Cooperation & Joint Collaboration </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.6;"> This project is built in close collaboration and joint effort with engineer <b>Kyle Hessling</b> to achieve stable fine-tuning of our 27B parameter model. <div style="margin-top: 10px; display: flex; align-items: center; gap: 6px;"> <span>👉</span> <span>You can follow him for hardware and model training updates on X / Twitter: <a href="https://x.com/KyleHessling1" target="_blank" style="color: #047857; text-decoration: none; font-weight: 700;">@KyleHessling1</a></span> </div> </div> </div> <!-- Card 3: Fine-tuning Framework --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #7c3aed 0%, #6d28d9 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>🦥</span> 1.3 Fine-tuning Framework (Unsloth) </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.6;"> Our model training is accelerated by the cool library from <b>Unsloth</b>. Special thanks to the Unsloth team for their highly efficient, memory-optimized training framework. <div style="margin-top: 10px; display: flex; align-items: center; gap: 6px;"> <span>👉</span> <span>You can visit their documentation for more fine-tuning guidance: <a href="https://unsloth.ai/docs" target="_blank" style="color: #7c3aed; text-decoration: none; font-weight: 700;">unsloth.ai/docs</a></span> </div> </div> </div> <!-- Card 4: Custom MTP Heads Processing & Automation Tooling --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #0ea5e9 0%, #0369a1 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>⚙️</span> 1.4 Custom MTP Heads Processing & Automation Tooling </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> This release features a custom splitting and merging methodology designed specifically for Qwen series Multi-Token Prediction (MTP) heads. The automation skill and complete processing pipeline scripts are open-sourced in <a href="https://github.com/R6410418/Jackrong-llm-finetuning-guide/tree/main/qwen-mtp-gguf" target="_blank" style="color: #0284c7; text-decoration: none; font-weight: 700;">qwen-mtp-gguf</a>. <div style="margin-top: 10px; display: flex; align-items: center; gap: 6px; color: #475569; font-weight: 500;"> <span>🌟</span><span>If you find this toolkit helpful, please support the project by leaving a star on GitHub!</span> </div> </div> </div> </div>[!TIP] Vision & Tool Calling Support: Qwopus3.6-27B-v2 natively supports vision and tool-use capabilities. To enable vision functionality, download
mmproj.gguffrom the GGUF Repository and place it in the same directory as the main.gguffile.
[!WARNING] Community Release Notice: Qwopus3.6-27B-v2 is an experimental community release and has not undergone complete safety evaluations or standard benchmarking. It is intended solely for research and exploration.
📖 2. Background & Motivation
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; display: flex; flex-direction: column; gap: 20px; margin-bottom: 30px;"> <!-- Card 2.1: The Distillation Dilemma --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #1e3a8a 0%, #1e40af 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>⚠️</span> 2.1 The Distillation Dilemma of "Reasoning Bubbles" </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.6;"> In the field of reasoning distillation, commercial closed-source models (such as GPT-4o or Claude 3.5 Sonnet) typically hide their raw step-by-step thinking processes, displaying only highly compressed "Reasoning Bubbles" to users. Direct imitation of these summaries introduces a severe <b>"Information Entropy Trap"</b>—student models struggle to mimic jumpy, step-skipped conclusions without undergoing the underlying logic derivations, resulting in reasoning fractures and poor generalizability. </div> </div> <!-- Card 2.2: Trace Inversion --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #0284c7 0%, #0369a1 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"> <span>🧬</span> 2.2 Trace Inversion & Negentropy Reconstruction </div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.6;"> To address this challenge, we introduce <b>Trace Inversion</b> datasets to reconstruct the full reasoning pathway. By using a specialized logical reconstructor, <a href="https://huggingface.co/Jackrong/Trace-Inverter-4B" target="_blank" style="color: #0369a1; text-decoration: none; font-weight: bold;">Trace-Inverter-4B</a>, we reverse-engineer the compressed reasoning bubbles into a complete, step-by-step <b>Learnable Chain-of-Thought (CoT)</b>. <br><br> This model integrates: <ul> <li><b>claude-opus-4.6-traceInversion-9000x</b>: 9,000 high-value, fully reconstructed step-by-step reasoning trajectories.</li> <li><b>claude-opus-4.7-traceInversion-5000x</b>: 5,000 complex multi-turn logic and mathematics samples optimized for negative entropy reconstruction.</li> </ul> This ensures the student model learns continuous, rigorous logical derivations rather than shortcut-ridden summaries. </div> </div> </div>⚡ 3. Reasoning Efficiency & MTP Speedup
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; border: 1px solid #cbd5e1; border-radius: 16px; box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.05), 0 4px 6px -2px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff; margin-bottom: 30px;"> <div style="background: linear-gradient(135deg, #0f766e 0%, #0e7490 100%); padding: 20px; color: white;"> <h3 style="margin: 0; font-size: 20px; font-weight: 700; display: flex; align-items: center; gap: 8px; color: white; border: none;">⚡ Reasoning Efficiency & MTP Acceleration</h3> <p style="margin: 4px 0 0 0; font-size: 13px; color: #ccfbf1;">A compact view of how many output tokens are needed to produce correct answers, and how the MTP variant improves inference throughput.</p> </div> <div style="padding: 24px; display: flex; flex-direction: column; gap: 24px;"> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(180px, 1fr)); gap: 16px;"> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #0f766e; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">🚀 MTP Speedup</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">1.66x</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">Qwopus3.6-27B-v2-MTP reaches 1.66x the official Qwen3.6 speed</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #0f766e; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">🧠 Correct-Answer Tokens</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">918.7</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">Average tokens when Qwopus answers correctly</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #0f766e; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">📉 Total Token Reduction</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">15.0%</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">627,325 vs 738,238 output tokens</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #0f766e; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">✅ Token Conversion</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">4.64</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">Correct answers per 10,000 output tokens</span> </div> </div> <div style="background: #f0fdfa; border-left: 4px solid #0f766e; border-radius: 0 8px 8px 0; padding: 12px 16px; font-size: 13px; color: #134e4a; line-height: 1.6;"> <b>Efficiency framing:</b> These measurements are based on parsed outputs from the same 350-question MMLU-style evaluation set. The goal is not only to compare accuracy, but also to measure how much reasoning text each model spends to reach correct answers. </div> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>🎯</span> 3.1 Correct-Answer Token Cost </div> <div style="padding: 16px;"> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 13px; min-width: 620px;"> <thead> <tr style="background: rgba(15, 118, 110, 0.06);"> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: left; color: #0f766e; font-weight: bold;">Metric Definition</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #475569; font-weight: bold;">Qwen3.6-27B</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Qwopus3.6-27B-v2</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Efficiency Gain</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);"><b>Definition A:</b> average output tokens on correctly answered questions only.</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">1,433.3 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #0f766e;">918.7 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #10b981;">35.9% fewer tokens</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);"><b>Definition B:</b> total output tokens divided by the number of correct answers, including token cost from wrong answers.</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">2,511.0 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #0f766e;">2,155.8 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #10b981;">14.2% less systemic overhead</td> </tr> </tbody> </table> </div> </div> </div> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>📈</span> 3.2 Token Conversion Efficiency </div> <div style="padding: 16px;"> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 13px; min-width: 620px; margin-bottom: 12px;"> <thead> <tr style="background: rgba(15, 118, 110, 0.06);"> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: left; color: #0f766e; font-weight: bold;">Metric</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #475569; font-weight: bold;">Qwen3.6-27B</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Qwopus3.6-27B-v2</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Delta</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: 600;">Correct answers per 10,000 output tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">3.98</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #0f766e;">4.64</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #10b981;">+16.6%</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: 600;">Total output token cost</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">738,238 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #0f766e;">627,325 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #10b981;">15.0% fewer tokens</td> </tr> </tbody> </table> </div> </div> </div> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>🧩</span> 3.3 Chain-of-Thought Length Comparison </div> <div style="padding: 16px;"> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 13px; min-width: 620px;"> <thead> <tr style="background: rgba(15, 118, 110, 0.06);"> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: left; color: #0f766e; font-weight: bold;">CoT Extraction Mode</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #475569; font-weight: bold;">Qwen3.6-27B</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Qwopus3.6-27B-v2</th> <th style="padding: 8px 10px; border-bottom: 2px solid #0f766e; text-align: right; color: #0f766e; font-weight: bold;">Reduction</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);"><b>Normal thinking endings:</b> text before the closing <code></think></code> tag only.</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">1,680.3 tokens<br><span style="font-size: 11px; color: #64748b;">5,169.4 chars</span></td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #0f766e;">798.5 tokens<br><span style="font-size: 11px; color: #64748b;">2,370.0 chars</span></td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #10b981;">52.5% shorter</td> </tr> </tbody> </table> </div> </div> </div> </div> </div>📊 4. Evaluation & Benchmarks
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; border: 1px solid #cbd5e1; border-radius: 16px; box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.05), 0 4px 6px -2px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff; margin-bottom: 30px;"> <div style="background: linear-gradient(135deg, #7c3aed 0%, #4f46e5 100%); padding: 20px; color: white;"> <h3 style="margin: 0; font-size: 20px; font-weight: 700; display: flex; align-items: center; gap: 8px; color: white; border: none;">📊 Evaluation & Performance Metrics</h3> <p style="margin: 4px 0 0 0; font-size: 13px; color: #ddd6fe;">Detailed benchmark results on MMLU-Pro, SWE-bench, frontend page layout generation, creative coding, and agentic reasoning.</p> </div> <div style="padding: 24px; display: flex; flex-direction: column; gap: 24px;"> <!-- Key Stats grid --> <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(180px, 1fr)); gap: 16px;"> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #7c3aed; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">📚 MMLU-Pro Subset</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">87.43%</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">+2.57 pp vs Qwen3.6-27B</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #7c3aed; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">🏆 SWE-bench Verified</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">75.25%</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">152 / 202 Resolved (Win)</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #7c3aed; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">🎨 Web Design</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">100%</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">5 / 5 Pages Validated</span> </div> <div style="border: 1px solid #e2e8f0; padding: 16px; border-radius: 10px; background: #f8fafc; text-align: center; box-shadow: inset 0 2px 4px rgba(0,0,0,0.02);"> <span style="font-size: 11px; font-weight: 700; color: #7c3aed; text-transform: uppercase; display: block; margin-bottom: 6px; letter-spacing: 0.5px;">⚡ Speed (RTX 5090)</span> <span style="font-size: 24px; font-weight: 800; color: #1e293b; display: block;">43.9</span> <span style="font-size: 11px; color: #64748b; font-weight: 500;">Avg tok/s (Q5_K_M)</span> </div> </div> <!-- Section 1: MMLU-Pro --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>📚</span> 4.1 MMLU-Pro Selected Subset </div> <div style="padding: 16px; display: flex; flex-direction: column; gap: 16px;"> <div style="background: #f8fafc; border: 1px solid #e2e8f0; border-radius: 8px; padding: 12px 14px; font-size: 13px; color: #334155; line-height: 1.6;"> <b>Evaluation format:</b> All models in this MMLU-Pro comparison were tested as unquantized, full-format 16-bit checkpoints. The selected evaluation subset covers 7 categories, with 50 questions per category and 350 questions in total. </div> <div style="overflow-x: auto; width: 100%;"> <table style="display: table; width: 100%; min-width: 100%; table-layout: fixed; border-collapse: collapse; font-size: 13px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="width: 42%; padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-weight: bold;">Model</th> <th style="width: 29%; padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Correct / Total</th> <th style="width: 29%; padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Accuracy</th> </tr> </thead> <tbody> <tr style="background: rgba(16, 185, 129, 0.06);"> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: 700; color: #047857;">Qwopus3.6-27B-v2</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 700; color: #047857;">306 / 350</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: 800; color: #10b981; font-size: 14px;">87.43%</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: 600;">Qwen3.6-27B</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">297 / 350</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold;">84.86%</td> </tr> </tbody> </table> </div> <div style="overflow-x: auto; width: 100%;"> <table style="display: table; width: 100%; min-width: 100%; table-layout: fixed; border-collapse: collapse; font-size: 12px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.03);"> <th style="width: 28%; padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: left;">Category</th> <th style="width: 24%; padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">Qwen3.6-27B</th> <th style="width: 28%; padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; color: #7c3aed;">Qwopus3.6-27B-v2</th> <th style="width: 20%; padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">Delta</th> </tr> </thead> <tbody> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Biology</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">96%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #7c3aed;">96%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; color: #64748b;">0 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Business</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">88%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">94%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">+6 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Computer Science</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">82%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">84%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">+2 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Mathematics</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">90%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #7c3aed;">88%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; color: #dc2626;">-2 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Physics</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">76%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">86%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">+10 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Chemistry</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">74%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">80%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #10b981;">+6 pp</td> </tr> <tr> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Health</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right;">88%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: 700; color: #7c3aed;">84%</td> <td style="padding: 7px 8px; border-bottom: 1px solid #e2e8f0; text-align: right; color: #dc2626;">-4 pp</td> </tr> </tbody> </table> </div> <p style="margin: 0; font-size: 13px; color: #475569; line-height: 1.6;"> <b>Summary:</b> On the selected 350-question MMLU-Pro evaluation set, Qwopus3.6-27B-v2 achieved <b>87.43%</b> accuracy, outperforming Qwen3.6-27B at <b>84.86%</b>. Qwopus3.6-27B-v2 is stronger in Business, Computer Science, Physics, and Chemistry, while Qwen3.6-27B remains ahead in Mathematics and Health. </p> <div style="display: inline-flex; align-items: center; width: fit-content; background: #ffffff; border: 1px solid #e2e8f0; border-radius: 999px; padding: 6px 10px; font-size: 11px; color: #64748b; font-weight: 600; box-shadow: 0 1px 2px rgba(15,23,42,0.04);"> Scope note: due to limited resources, only randomly sampled questions from MMLU-Pro were evaluated. To keep the evaluation fair and transparent, model response logs will be placed in the repository test_data folder. </div> </div> </div> <!-- Section 2: SWE-bench --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>💻</span> 4.2 SWE-bench Verified (controlled-202 slice) </div> <div style="padding: 16px;"> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 13px; margin-bottom: 12px; min-width: 500px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-weight: bold;">Model / Configuration</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-weight: bold;">Sampling</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Resolved</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Empty Patches</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Resolve %</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: 700;">Qwopus 3.6 27B v2 (dense)</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15);">temp 1.0, step 275, single-slot</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold; color: #10b981;">152 / 202</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">1</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold; color: #10b981; font-size: 14px;">75.25%</td> </tr> </tbody> </table> </div> <p style="margin: 0 0 12px 0; font-size: 13px; color: #475569; line-height: 1.6;"> <b>Execution Details:</b> 19h 29m wall-clock on a single RTX 5090 using a 160K fp16 context window. Every instance successfully exited with <code>Submitted</code> status. 0 step-limit hits and 0 context-overflow failures occurred. Median trajectory length was 67 / 275 steps. </p> <div style="background: #fffbeb; border-left: 4px solid #d97706; border-radius: 0 8px 8px 0; padding: 12px 16px; font-size: 13px; color: #92400e; display: flex; gap: 10px; align-items: flex-start; line-height: 1.5;"> <span style="font-size: 16px; margin-top: 2px;">⚡</span> <div> <b>Run agentic harnesses hot:</b> Raising sampling temperature to <b>temp=1.0</b> with thinking-on effectively eliminates the reasoning-loop failure mode where earlier finetunes hit 78 empty patches. Greedy decoding (temp=0.1) forces the finetune to over-deliberate and loop inside the <code><think></code> block, whereas a higher temperature enables the model to utilize the full breadth of reasoning paths established during training. </div> </div> </div> </div> <!-- Section 3: Throughput Comparison --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>⚡</span> 4.3 Throughput & VRAM Allocation (RTX 5090) </div> <div style="padding: 16px;"> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 13px; margin-bottom: 12px; min-width: 500px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-weight: bold;">Metric</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #475569; font-weight: bold;">Qwopus 3.6 35B-A3B (MoE, Q5)</th> <th style="padding: 8px 10px; border-bottom: 2px solid #7c3aed; text-align: right; color: #7c3aed; font-weight: bold;">Qwopus 3.6 27B V2 (Dense, Q5)</th> </tr> </thead> <tbody> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">Average Throughput</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">161.9 tok/s</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold; color: #7c3aed;">43.9 tok/s</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">Throughput Range</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">154.4 / 164.8 tok/s</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">43.1 / 44.6 tok/s</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">VRAM Usage</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">~25 GB (65K q8 context)</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold;">~31 GB (160K fp16 context)</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">Completion Tokens (Suite)</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">106,688 tokens</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold;">119,036 tokens</td> </tr> <tr> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold;">Total Runtime (Suite)</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right;">11.1 min</td> <td style="padding: 8px 10px; border-bottom: 1px solid rgba(128,128,128,0.15); text-align: right; font-weight: bold;">45.3 min</td> </tr> </tbody> </table> </div> <p style="margin: 0; font-size: 13px; color: #475569; line-height: 1.6;"> <b>Architecture Trade-off:</b> The MoE wins on raw throughput by ~3.7x due to its A3B routing pattern. However, the Dense 27B model offsets this with superior per-token reasoning depth. We recommend the <b>Dense 27B model</b> for complex agentic workflows, long-context reasoning, and code execution, and the <b>MoE model</b> for fast, high-throughput generations. Tight throughput variance (±0.75 tok/s) indicates the dense model is fully memory-bandwidth-bound. </p> </div> </div> <!-- Section 4: Detailed Tasks --> <div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%); padding: 12px 16px; border-bottom: 1px solid #cbd5e1; font-weight: 700; font-size: 14px; color: #1e293b; display: flex; align-items: center; gap: 8px;"> <span>🎨</span> 4.4 Web Design, WebGL Canvas, and Agentic Task Breakdown </div> <div style="padding: 16px; display: flex; flex-direction: column; gap: 20px;"> <!-- Web Design Layouts --> <div> <span style="font-weight: 700; font-size: 13px; color: #334155; display: block; margin-bottom: 8px; border-bottom: 1px dashed #cbd5e1; padding-bottom: 4px;">🎨 Web Design Layout Generation (All 5 validated end-to-end)</span> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 12px; min-width: 480px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.03);"> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: left;">Prompt / Brief</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Size (KB)</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Tokens</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Time</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Reasoning Tokens</th> </tr> </thead> <tbody> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">SaaS Landing Page (AI Observability)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">60.3 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">23,801</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">552 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: bold; color: #7c3aed;">836</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Analytics Dashboard (Light Theme)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">42.1 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">15,390</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">354 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: bold; color: #7c3aed;">1,898</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Designer Portfolio (Kinetic Typography)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">32.5 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">11,612</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">265 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: bold; color: #7c3aed;">1,459</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Pricing Page (3 Tiers + FAQ)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">26.6 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">9,360</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">213 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: bold; color: #7c3aed;">1,077</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Mobile App Marketing Page</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">42.3 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">16,590</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">382 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; font-weight: bold; color: #7c3aed;">1,650</td> </tr> </tbody> </table> </div> </div> <!-- WebGL Canvas Sketches --> <div> <span style="font-weight: 700; font-size: 13px; color: #334155; display: block; margin-bottom: 8px; border-bottom: 1px dashed #cbd5e1; padding-bottom: 4px;">🌌 Canvas / WebGL Creative Coding (Selected Publications)</span> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 12px; min-width: 480px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.03);"> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: left;">Sketch Name</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Size (KB)</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Tokens</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Time</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: left; padding-left: 12px;">Configuration & Metrics</th> </tr> </thead> <tbody> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Particle Attractor (Fluid Swarm)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">9.4 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">4,308</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">97 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; padding-left: 12px; color: #475569;">temp 1.0 · 1,513 chars reasoning</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Generative Flowfield (Ink Agents)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">13.9 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">7,237</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">163 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; padding-left: 12px; color: #475569;">temp 1.0 · 6,269 chars reasoning</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Soft-Body Physics Sandbox</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">18.0 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">6,827</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">154 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; padding-left: 12px; color: #475569;">temp 0.75 · 1,665 chars reasoning (shipped clean first run)</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Audio-Reactive Visualizer</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">10.7 KB</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">5,731</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">129 s</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; padding-left: 12px; color: #475569;">temp 1.0 · 7,645 chars reasoning</td> </tr> </tbody> </table> </div> <span style="font-size: 11px; color: #64748b; margin-top: 4px; display: block;">*Note: Mandelbulb and Three.js crystal scene outputs are excluded due to aesthetic standards and parked in <code>excluded-canvas/</code>.</span> </div> <!-- Agentic Reasoning --> <div> <span style="font-weight: 700; font-size: 13px; color: #334155; display: block; margin-bottom: 8px; border-bottom: 1px dashed #cbd5e1; padding-bottom: 4px;">🧠 Agentic Tasks (5 prompts + 1 structured extraction nothink rerun)</span> <div style="overflow-x: auto;"> <table style="width: 100%; border-collapse: collapse; font-size: 12px; min-width: 480px; margin-bottom: 8px;"> <thead> <tr style="background: rgba(124, 58, 237, 0.03);"> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: left;">Task Brief</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Completion Tokens</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Reasoning Characters</th> <th style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">Time</th> </tr> </thead> <tbody> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Multi-step Planning (URL shortener deploy)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">2,238</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">7,067</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">50 s</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Tool-use Planning (Flights, Hotel, Weather)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">1,262</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">2,807</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">28 s</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Code Debugging (4-bug BST K-th smallest)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">1,753</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">5,225</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">39 s</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Structured Extraction (Roster from prose)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">1,721</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">4,245</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">39 s</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Self-Critique Loop (Palindrome optimization)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">1,255</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">3,309</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">28 s</td> </tr> <tr> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; font-weight: 600;">Structured Extraction (No-think)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">351</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right; color: #94a3b8;">0 (nothink)</td> <td style="padding: 6px; border-bottom: 1px solid #e2e8f0; text-align: right;">8 s</td> </tr> </tbody> </table> </div> <ul style="margin: 0; padding-left: 18px; font-size: 12px; color: #475569; line-height: 1.6; display: flex; flex-direction: column; gap: 4px;"> <li><b>code_debug:</b> Successfully caught all 4 bugs (sort order, <code>=</code> vs <code>==</code>, useless loop, off-by-one errors).</li> <li><b>self_critique:</b> Followed the structured instruction loop (INITIAL → CRITIQUE → IMPROVED) and optimized a palindrome algorithm to O(n²) expand-around-center.</li> <li><b>multi_step_planning:</b> Designed a robust 10-step deployment plan with Dockerfile hand-off and explicit pip dependencies.</li> <li><b>tool_use_json:</b> Resolved a 3-tool sequence (<code>search_flights</code>, <code>book_hotel</code>, <code>get_weather</code>) with completely valid argument shapes.</li> </ul> </div> </div> </div> </div> </div>🗺️ 5. Training & Data Pipeline Overview
The training process fuses Trace Inversion data augmentation with a Three-Stage Curriculum Learning pipeline. The core engineering focuses on expanding context length gradually while training on reconstructed reasoning traces to guarantee format stability.
[ 🗺️ Trace Inversion: Reconstructing Distillation Workflow ]
A. Surrogate Model Training (Trace Inverter)
Open-source Model (GLM-5.1 / DS-V4) ──► Complete Reasoning Chain ──► [ Qwen3-235B Compression ] ──► Reasoning Bubbles
│ │
└──────────► [ Training ] ◄─────────┘
(Base: Qwen3-4B-Instruct)
(Result: Trace-Inverter-4B)
B. Inversion Phase: Reconstructing Claude-4.7-Max
_______________________________________________________
| |
| Claude-4.7-Max API ──► Compressed Bubbles + Answer |
|_______________________________________________________|
│
▼
[ 🧠 Trace-Inverter-4B (Logic Reconstructor) ] ──► Synthetic Deep Reasoning Trace (Learnable CoT)
│
▼
[ 🧩 Data Splicing ] ◄────────── (Original Prompt + Response)
(Embed reconstructed CoT in <think> tags, splicing with original prompt/response)
│
▼
(Result: claude-opus-4.6/4.7 inverted sets)
C. Final SFT Curriculum Pipeline
___________________________________________
| |
| Base Model (Qwen3.6-27B) |
|___________________________________________|
│
▼
[ 📦 Phase 1: Format Inception ] ──► [ 🛠️ Phase 2: Complexity Expansion ] ──► [ 🚀 Phase 3: Long-Context SFT ]
( < 4096 tokens ) ( 4096 - 8192 tokens ) ( 8192 - 32K tokens )
(Short-context stable format) (Medium-complexity reasoning) (Long/Multi-turn / 10% replay)
│ │
└─────────────────────────────┬─────────────────────────────────────────┘
▼
_____________________________________________
| |
| 🌟 Final Model: Qwopus3.6-27B-v2 |
|_____________________________________________|
🎯 6. Three-Stage Curriculum Learning
To steadily scale up the reasoning quality under long-context inference, Qwopus3.6-27B-v2 adopts a Curriculum Learning strategy, progressively mixing longer and more complex reasoning templates:
<table style="width: 100%; border-collapse: collapse; margin-top: 15px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="padding: 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-size: 14px; width: 25%;">Curriculum Stage</th> <th style="padding: 10px; border-bottom: 2px solid #7c3aed; text-align: left; font-size: 14px; width: 35%;">Focus & Sample Characteristics</th> <th style="padding: 10px; border-bottom: 2px solid #7c3aed; text-align: left; font-size: 14px;">Strategy Details</th> </tr> </thead> <tbody> <tr> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px; color: #7c3aed;">📦 Stage 1: Format Inception</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">• Limit context within 4,096 tokens<br>• Emphasize stable reasoning templates</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">Focuses on short-to-medium length, cleanly formatted reasoning samples. The primary goal is to establish a reliable, structured reasoning output format (such as auto-closing <code><think></code> tags), preventing premature exposure to complex chains from causing format collapse.</td> </tr> <tr> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px; color: #7c3aed;">🛠️ Stage 2: Complexity Expansion</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">• Extend length to 4,096 - 8,192 tokens<br>• Introduce high-difficulty logic samples</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">Gradually increases the ratio of complex reasoning chains. By aligned distillation with "teacher models" whose reasoning style distributions closely match the Qwen3.6 base, the capacity gap is controlled to achieve highly efficient knowledge transfer.</td> </tr> <tr> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px; color: #7c3aed;">🚀 Stage 3: Long-Context SFT</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">• Progressively scale window up to 32K tokens<br>• 10% high-quality short sample replay</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">In this stage, the model is pushed to deep reasoning scenarios under ultra-long context and multi-turn dialogues. To prevent capacity drift or degradation of short-instruction comprehension during long-text training, a 10% replay of high-quality short samples is strictly enforced.</td> </tr> </tbody> </table>🎨 7. Trace Inversion Case Studies (5 Key Domains Showcase)
To demonstrate how Trace Inversion reconstructs logical continuity and eliminates negative entropy, the following interactive panels show the contrast between raw compressed "Reasoning Bubbles" and the fully step-by-step reconstructed chain-of-thought (Learnable CoT) under 5 typical scenarios:
📐 Domain 1: Mathematics (Probability Calculation)
<!-- Math Case --> <div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif; margin-bottom: 30px; border: 1px solid #cbd5e1; border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff;"> <div style="background: linear-gradient(135deg, #7c3aed 0%, #4c1d95 100%); padding: 12px 16px; display: flex; align-items: center; justify-content: space-between;"> <div style="display: flex; align-items: center; gap: 8px;"> <span style="font-size: 16px;">📐</span> <span style="color: #ffffff; font-weight: 700; font-size: 14px;">Case 1: Probability Calculation</span> </div> <span style="background: rgba(255, 255, 255, 0.2); color: #ffffff; font-size: 11px; font-weight: 600; padding: 2px 8px; border-radius: 12px;">Math Case</span> </div> <div style="display: flex; flex-wrap: wrap; padding: 16px; gap: 16px; background: #f8fafc;"> <div style="flex: 1; min-width: 280px; display: flex; flex-direction: column; gap: 12px;"> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #7c3aed; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💬 Query</div> <div style="padding: 10px; font-size: 12px; color: #1e293b; line-height: 1.5;"> A bag contains 4 red balls and 6 blue balls. If two balls are drawn at random without replacement, what is the probability that both are red? </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #6d28d9; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">📋 Compressed Reasoning Bubble</div> <div style="padding: 10px; font-size: 12px; color: #475569; line-height: 1.5; font-style: italic;"> First draw red: 4/10. Second draw red: 3/9. Multiply: (4/10) * (3/9) = 12/90 = 2/15. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #4c1d95; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">✅ Final Answer</div> <div style="padding: 10px; font-size: 12px; color: #0f172a; line-height: 1.5;"> The probability that both balls drawn are red is 2/15.<div style="margin-top: 6px; padding: 6px; background: #f5f3ff; border: 1px solid #ddd6fe; border-radius: 4px; color: #6d28d9; font-weight: 700; text-align: center;">2/15</div> </div> </div> </div> <div style="flex: 1.2; min-width: 320px; border: 1px solid #a7f3d0; border-radius: 8px; background: #ffffff; overflow: hidden; display: flex; flex-direction: column;"> <div style="background: #065f46; color: #ffffff; padding: 8px 12px; font-size: 12px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💡 Synthesized Reasoning Trace (Learnable CoT)</div> <div style="padding: 12px; display: flex; flex-direction: column; gap: 10px; flex-grow: 1; font-size: 12px; color: #334155; line-height: 1.5;"> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 1<br><span style='font-weight: 500; color: #047857;'>Define</span> </div> <div>Identify total items and counts: Total balls = 4 (red) + 6 (blue) = 10 balls. The goal is to compute the probability of drawing two red balls in succession without replacement: P(R1 ∩ R2).</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 2<br><span style='font-weight: 500; color: #047857;'>First Draw</span> </div> <div>Calculate probability of the first event (drawing a red ball first): P(R1) = Number of red balls / Total balls = 4/10. This simplifies to 2/5.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 3<br><span style='font-weight: 500; color: #047857;'>Second Draw</span> </div> <div>Calculate probability of the second event given that the first was red: Since there is no replacement, there are now 9 balls left in the bag, and 3 of them are red. P(R2 | R1) = 3/9 = 1/3.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 4<br><span style='font-weight: 500; color: #047857;'>Multiply</span> </div> <div>Apply the multiplication rule for dependent events: P(R1 ∩ R2) = P(R1) × P(R2 | R1). Substitute values: P(R1 ∩ R2) = (4/10) × (3/9) = 12/90.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 5<br><span style='font-weight: 500; color: #047857;'>Simplify</span> </div> <div>Simplify the resulting fraction: 12/90 can be divided by 6 in both numerator and denominator: 12÷6 = 2, 90÷6 = 15. The final probability is 2/15.</div> </div> <div style="margin-top: auto; display: flex; gap: 8px; background: #f0fdf4; border: 1px solid #a7f3d0; border-radius: 6px; padding: 8px 12px; align-items: center;"> <span style="font-size: 14px; color: #047857;">🏆</span> <div style="font-size: 12px; font-weight: 700; color: #065f46;"> Conclusion: The probability of drawing both red balls is 2/15 (~13.33%). </div> </div> </div> </div> </div> </div>🚀 Domain 2: Physics (Kinematics)
<!-- Physics Case --> <div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif; margin-bottom: 30px; border: 1px solid #cbd5e1; border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff;"> <div style="background: linear-gradient(135deg, #1e3a8a 0%, #1e40af 100%); padding: 12px 16px; display: flex; align-items: center; justify-content: space-between;"> <div style="display: flex; align-items: center; gap: 8px;"> <span style="font-size: 16px;">🚀</span> <span style="color: #ffffff; font-weight: 700; font-size: 14px;">Case 2: Kinematics Formula</span> </div> <span style="background: rgba(255, 255, 255, 0.2); color: #ffffff; font-size: 11px; font-weight: 600; padding: 2px 8px; border-radius: 12px;">Physics Case</span> </div> <div style="display: flex; flex-wrap: wrap; padding: 16px; gap: 16px; background: #f8fafc;"> <div style="flex: 1; min-width: 280px; display: flex; flex-direction: column; gap: 12px;"> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1e3a8a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💬 Query</div> <div style="padding: 10px; font-size: 12px; color: #1e293b; line-height: 1.5;"> A car starts from rest and accelerates uniformly at 3 m/s² for 6 seconds. How far does the car travel during this time? </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1b497a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">📋 Compressed Reasoning Bubble</div> <div style="padding: 10px; font-size: 12px; color: #475569; line-height: 1.5; font-style: italic;"> Use kinematics equation d = v0*t + 0.5*a*t^2. Initial velocity v0 = 0, a = 3, t = 6. Calculate 0.5 * 3 * 36. Output 54 meters. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #0369a1; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">✅ Final Answer</div> <div style="padding: 10px; font-size: 12px; color: #0f172a; line-height: 1.5;"> The car travels 54 meters.<div style="margin-top: 6px; padding: 6px; background: #f0f9ff; border: 1px solid #bae6fd; border-radius: 4px; color: #0369a1; font-weight: 700; text-align: center;">Answer: 54 meters</div> </div> </div> </div> <div style="flex: 1.2; min-width: 320px; border: 1px solid #a7f3d0; border-radius: 8px; background: #ffffff; overflow: hidden; display: flex; flex-direction: column;"> <div style="background: #065f46; color: #ffffff; padding: 8px 12px; font-size: 12px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💡 Synthesized Reasoning Trace (Learnable CoT)</div> <div style="padding: 12px; display: flex; flex-direction: column; gap: 10px; flex-grow: 1; font-size: 12px; color: #334155; line-height: 1.5;"> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 1<br><span style='font-weight: 500; color: #047857;'>Goal</span> </div> <div>Identify the given variables and goal: Calculate the displacement (d) traveled under uniform acceleration from rest.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 2<br><span style='font-weight: 500; color: #047857;'>Extract</span> </div> <div>Identify given values: Initial velocity (v<sub>0</sub>) = 0 m/s (starts from rest), acceleration (a) = 3 m/s², and duration (t) = 6 seconds.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 3<br><span style='font-weight: 500; color: #047857;'>Match Formula</span> </div> <div>Select the kinematic formula relating displacement to acceleration and time: d = v<sub>0</sub>t + ½at<sup>2</sup>.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 4<br><span style='font-weight: 500; color: #047857;'>Compute</span> </div> <div>Substitute the values: d = (0)×(6) + ½×3×(6)<sup>2</sup>. Simplify: first term is 0; 6<sup>2</sup> = 36; 3 × 36 = 108; ½ × 108 = 54 meters.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 5<br><span style='font-weight: 500; color: #047857;'>Verify</span> </div> <div>Verify using average velocity method: final velocity v<sub>final</sub> = v<sub>0</sub> + at = 0 + 3×6 = 18 m/s. Average velocity v<sub>avg</sub> = (0 + 18)/2 = 9 m/s. Distance d = v<sub>avg</sub> × t = 9 × 6 = 54 meters. The calculation is robust.</div> </div> <div style="margin-top: auto; display: flex; gap: 8px; background: #f0fdf4; border: 1px solid #a7f3d0; border-radius: 6px; padding: 8px 12px; align-items: center;"> <span style="font-size: 14px; color: #047857;">🏆</span> <div style="font-size: 12px; font-weight: 700; color: #065f46;"> Conclusion: The car travels 54 meters during this uniform acceleration period. </div> </div> </div> </div> </div> </div>💻 Domain 3: Coding (Algorithm Logic)
<!-- Programming Case --> <div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif; margin-bottom: 30px; border: 1px solid #cbd5e1; border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff;"> <div style="background: linear-gradient(135deg, #1e3a8a 0%, #1e40af 100%); padding: 12px 16px; display: flex; align-items: center; justify-content: space-between;"> <div style="display: flex; align-items: center; gap: 8px;"> <span style="font-size: 16px;">💻</span> <span style="color: #ffffff; font-weight: 700; font-size: 14px;">Case 3: Algorithm Logic</span> </div> <span style="background: rgba(255, 255, 255, 0.2); color: #ffffff; font-size: 11px; font-weight: 600; padding: 2px 8px; border-radius: 12px;">Coding Case</span> </div> <div style="display: flex; flex-wrap: wrap; padding: 16px; gap: 16px; background: #f8fafc;"> <div style="flex: 1; min-width: 280px; display: flex; flex-direction: column; gap: 12px;"> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1e3a8a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💬 Query</div> <div style="padding: 10px; font-size: 12px; color: #1e293b; line-height: 1.5;"> Given an array of integers, find the sum of all even numbers. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1b497a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">📋 Compressed Reasoning Bubble</div> <div style="padding: 10px; font-size: 12px; color: #475569; line-height: 1.5; font-style: italic;"> Filter evens using x % 2 == 0, then sum them up. Iterate through array, checking modulus. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #0369a1; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">✅ Final Answer</div> <div style="padding: 10px; font-size: 12px; color: #0f172a; line-height: 1.5;"> Here is the Python implementation:<pre style="margin: 6px 0 0 0; background: #f1f5f9; padding: 8px; border-radius: 4px; font-family: monospace; font-size: 11px; color: #0f172a; overflow-x: auto; border: 1px solid #cbd5e1; line-height: 1.4;">def sum_even_numbers(arr): return sum(x for x in arr if x % 2 == 0)</pre> </div> </div> </div> <div style="flex: 1.2; min-width: 320px; border: 1px solid #a7f3d0; border-radius: 8px; background: #ffffff; overflow: hidden; display: flex; flex-direction: column;"> <div style="background: #065f46; color: #ffffff; padding: 8px 12px; font-size: 12px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💡 Synthesized Reasoning Trace (Learnable CoT)</div> <div style="padding: 12px; display: flex; flex-direction: column; gap: 10px; flex-grow: 1; font-size: 12px; color: #334155; line-height: 1.5;"> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 1<br><span style='font-weight: 500; color: #047857;'>Iterate</span> </div> <div>Initialize a tracking sum variable to 0. Prepare to iterate through the array from index 0 to N-1.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 2<br><span style='font-weight: 500; color: #047857;'>Check Even</span> </div> <div>For each element x, apply the modulus operator: check if x % 2 == 0. If True, the number is even.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 3<br><span style='font-weight: 500; color: #047857;'>Accumulate</span> </div> <div>If the condition is met, add x to the tracking sum. Otherwise, proceed to the next iteration.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 4<br><span style='font-weight: 500; color: #047857;'>Edge Cases</span> </div> <div>Check for empty arrays or arrays with no even numbers. In both cases, the returned sum is 0. Negative numbers are handled correctly (e.g. -2 % 2 == 0 is True).</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 5<br><span style='font-weight: 500; color: #047857;'>Complexity</span> </div> <div>Verify complexity: The algorithm runs in O(N) time as it inspects each element exactly once, and O(1) auxiliary space.</div> </div> <div style="margin-top: auto; display: flex; gap: 8px; background: #f0fdf4; border: 1px solid #a7f3d0; border-radius: 6px; padding: 8px 12px; align-items: center;"> <span style="font-size: 14px; color: #047857;">🏆</span> <div style="font-size: 12px; font-weight: 700; color: #065f46;"> Conclusion: Summing even numbers completed with O(N) time complexity. </div> </div> </div> </div> </div> </div>🧠 Domain 4: Logical Reasoning (Syllogism)
<!-- Logical Reasoning Case --> <div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif; margin-bottom: 30px; border: 1px solid #cbd5e1; border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff;"> <div style="background: linear-gradient(135deg, #1e3a8a 0%, #1e40af 100%); padding: 12px 16px; display: flex; align-items: center; justify-content: space-between;"> <div style="display: flex; align-items: center; gap: 8px;"> <span style="font-size: 16px;">🧠</span> <span style="color: #ffffff; font-weight: 700; font-size: 14px;">Case 4: Syllogism & Set Inclusion</span> </div> <span style="background: rgba(255, 255, 255, 0.2); color: #ffffff; font-size: 11px; font-weight: 600; padding: 2px 8px; border-radius: 12px;">Logic Case</span> </div> <div style="display: flex; flex-wrap: wrap; padding: 16px; gap: 16px; background: #f8fafc;"> <div style="flex: 1; min-width: 280px; display: flex; flex-direction: column; gap: 12px;"> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1e3a8a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💬 Query</div> <div style="padding: 10px; font-size: 12px; color: #1e293b; line-height: 1.5;"> If all philosophers are thinkers, and some thinkers are writers, is it necessarily true that some philosophers are writers? </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #1b497a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">📋 Compressed Reasoning Bubble</div> <div style="padding: 10px; font-size: 12px; color: #475569; line-height: 1.5; font-style: italic;"> Class inclusion. 'Philosophers' is a subset of 'Thinkers'. 'Writers' overlaps with 'Thinkers', but the overlap could lie entirely outside the 'Philosophers' subset. Thus, it does not necessarily follow. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #0369a1; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">✅ Final Answer</div> <div style="padding: 10px; font-size: 12px; color: #0f172a; line-height: 1.5;"> No, it does not necessarily follow.<div style="margin-top: 6px; padding: 6px; background: #f0f9ff; border: 1px solid #bae6fd; border-radius: 4px; color: #0369a1; font-weight: 700; text-align: center;">Answer: No, it does not follow</div> </div> </div> </div> <div style="flex: 1.2; min-width: 320px; border: 1px solid #a7f3d0; border-radius: 8px; background: #ffffff; overflow: hidden; display: flex; flex-direction: column;"> <div style="background: #065f46; color: #ffffff; padding: 8px 12px; font-size: 12px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💡 Synthesized Reasoning Trace (Learnable CoT)</div> <div style="padding: 12px; display: flex; flex-direction: column; gap: 10px; flex-grow: 1; font-size: 12px; color: #334155; line-height: 1.5;"> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 1<br><span style="font-weight: 500; color: #047857;">Set Definition</span> </div> <div>Define the sets: P = Philosophers, T = Thinkers, W = Writers. Transcribe premises: Premise 1 is P ⊆ T. Premise 2 is T ∩ W ≠ ∅.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 2<br><span style="font-weight: 500; color: #047857;">Objective</span> </div> <div>Determine if the conclusion P ∩ W ≠ ∅ is a logical necessity under all models.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 3<br><span style="font-weight: 500; color: #047857;">Venn Analysis</span> </div> <div>Visualize via Venn diagrams. Circle P is completely inside Circle T. Circle W intersects Circle T. W can intersect only the region of T that excludes P. In this layout, P ∩ W = ∅.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 4<br><span style="font-weight: 500; color: #047857;">Counterexample</span> </div> <div>Create a semantic counterexample: Let T = Humans, P = French people, W = Japanese people. All French people are Humans. Some Humans are Japanese. Does it follow that some French people are Japanese? No.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 5<br><span style="font-weight: 500; color: #047857;">Conclusion</span> </div> <div>Since a valid counterexample exists, the argument forms an invalid syllogism (specifically, Fallacy of the Undistributed Middle). The conclusion is not guaranteed.</div> </div> <div style="margin-top: auto; display: flex; gap: 8px; background: #f0fdf4; border: 1px solid #a7f3d0; border-radius: 6px; padding: 8px 12px; align-items: center;"> <span style="font-size: 14px; color: #047857;">🏆</span> <div style="font-size: 12px; font-weight: 700; color: #065f46;"> Conclusion: The syllogism is invalid; the conclusion does not necessarily follow. </div> </div> </div> </div> </div> </div>💡 Domain 5: Core Theory (Reasoning Bubble vs. Learnable CoT)
<!-- Conceptual Case --> <div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif; margin-bottom: 30px; border: 1px solid #cbd5e1; border-radius: 12px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05); overflow: hidden; background: #ffffff;"> <div style="background: linear-gradient(135deg, #0f172a 0%, #1e293b 100%); padding: 12px 16px; display: flex; align-items: center; justify-content: space-between;"> <div style="display: flex; align-items: center; gap: 8px;"> <span style="font-size: 16px;">💡</span> <span style="color: #ffffff; font-weight: 700; font-size: 14px;">Case 5: Proprietary API "Reasoning Bubbles" vs. Reconstructed "Learnable CoT"</span> </div> <span style="background: rgba(255, 255, 255, 0.2); color: #ffffff; font-size: 11px; font-weight: 600; padding: 2px 8px; border-radius: 12px;">Core Theory</span> </div> <div style="display: flex; flex-wrap: wrap; padding: 16px; gap: 16px; background: #f8fafc;"> <div style="flex: 1; min-width: 280px; display: flex; flex-direction: column; gap: 12px;"> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #0f172a; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💬 Core Problem</div> <div style="padding: 10px; font-size: 12px; color: #1e293b; line-height: 1.5;"> What is the difference between proprietary API "Reasoning Bubbles" and reconstructed "Learnable Chain-of-Thought (CoT)"? Why is direct fine-tuning on Reasoning Bubbles problematic? </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #334155; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">📋 Compressed Reasoning Bubble</div> <div style="padding: 10px; font-size: 12px; color: #475569; line-height: 1.5; font-style: italic;"> Reasoning bubbles are condensed summaries of thinking. They hide step-by-step logic, making direct imitation hard for small models. Reconstructed CoT fills in missing steps, providing a smooth learning gradient for model distillation. </div> </div> <div style="border: 1px solid #e2e8f0; border-radius: 8px; background: #ffffff; overflow: hidden;"> <div style="background: #0284c7; color: #ffffff; padding: 6px 10px; font-size: 11px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">✅ Final Answer</div> <div style="padding: 10px; font-size: 12px; color: #0f172a; line-height: 1.5;"> Proprietary "Reasoning Bubbles" serve as compressed, high-level summaries that cause cognitive gaps in student models. "Learnable CoT" is a step-by-step, fully-derived logical chain that restores intermediate gradients, enabling effective reasoning distillation. <div style="margin-top: 6px; padding: 6px; background: #f0f9ff; border: 1px solid #bae6fd; border-radius: 4px; color: #0369a1; font-weight: 700; text-align: center;"> Bubble = What was thought; CoT = How it was thought </div> </div> </div> </div> <div style="flex: 1.2; min-width: 320px; border: 1px solid #a7f3d0; border-radius: 8px; background: #ffffff; overflow: hidden; display: flex; flex-direction: column;"> <div style="background: #065f46; color: #ffffff; padding: 8px 12px; font-size: 12px; font-weight: 700; letter-spacing: 0.5px; text-transform: uppercase;">💡 Synthesized Reasoning Trace (Learnable CoT)</div> <div style="padding: 12px; display: flex; flex-direction: column; gap: 10px; flex-grow: 1; font-size: 12px; color: #334155; line-height: 1.5;"> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 1<br><span style="font-weight: 500; color: #047857;">Define</span> </div> <div><b>Define the Concepts:</b> Reasoning Bubbles are high-level summaries written for humans, whereas Learnable CoT consists of explicit, step-by-step derivations optimized for model learning.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 2<br><span style="font-weight: 500; color: #047857;">Entropy</span> </div> <div><b>Information Loss (Entropy Trap):</b> API summaries hide intermediate calculation steps. Direct imitation introduces severe "information entropy traps"—student models are forced to skip logical steps, causing logical confusion.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 3<br><span style="font-weight: 500; color: #047857;">Gradient</span> </div> <div><b>Causal Gradients:</b> Reconstructed CoT unfolds every derivation, calculation, and code design block. By providing continuous token-level supervision, it creates stable logical gradients for model weights to converge on.</div> </div> <div style="display: flex; gap: 10px; align-items: flex-start; border-bottom: 1px dashed #e2e8f0; padding-bottom: 8px;"> <div style="background: #ecfdf5; border: 1px solid #10b981; border-radius: 4px; padding: 4px; min-width: 65px; text-align: center; font-size: 9px; font-weight: 700; color: #065f46;"> STEP 4<br><span style="font-weight: 500; color: #047857;">Inversion</span> </div> <div><b>Trace Inversion Process:</b> Using the problem and the final answer as hard endpoints, and the reasoning bubble as clues, the inversion model reverse-engineers a complete, rigorous reasoning path (Learnable CoT).</div> </div> <div style="margin-top: auto; display: flex; gap: 8px; background: #f0fdf4; border: 1px solid #a7f3d0; border-radius: 6px; padding: 8px 12px; align-items: center;"> <span style="font-size: 14px; color: #047857;">🏆</span> <div style="font-size: 12px; font-weight: 700; color: #065f46;"> Conclusion: Reconstructed CoT turns dark black-box data into open, highly learnable logic gradients. </div> </div> </div> </div> </div> </div>🤝 8. Collaboration & Training Details
This model is a collaborative milestone achieved with hardware engineer Kyle Hessling. You can follow him on X / Twitter: @KyleHessling1 to keep up with the latest hardware infrastructure and distributed training updates. 🙏
<table style="width: 100%; border-collapse: collapse; margin-top: 15px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif;"> <thead> <tr style="background: rgba(124, 58, 237, 0.05);"> <th style="padding: 10px; border-bottom: 2px solid #7c3aed; text-align: left; color: #7c3aed; font-size: 14px; width: 30%;">Dimension</th> <th style="padding: 10px; border-bottom: 2px solid #7c3aed; text-align: left; font-size: 14px;">Details & Infrastructure</th> </tr> </thead> <tbody> <tr> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px;">🖥️ Training Hardware</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">NVIDIA DGX Cluster / H100 / RTX 6000 Pro</td> </tr> <tr> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-weight: bold; font-size: 13px;">⚙️ Fine-tuning Framework</td> <td style="padding: 10px; border-bottom: 1px solid rgba(128,128,128,0.15); font-size: 13px;">Unsloth (used for highly efficient SFT of dense models and memory optimization)</td> </tr> </tbody> </table>⚠️ 9. Known Training & Deployment Issues (IMPORTANT)
While the 27B dense model architecture is relatively stable, certain low-level framework compatibility issues may still surface during large-scale parameter updates and complex long-context training. It is highly recommended to monitor the following technical risk points during secondary fine-tuning and deployment:
<table style="width: 100%; border-collapse: collapse; margin-top: 15px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Helvetica, Arial, sans-serif;"> <thead> <tr style="background: rgba(220, 38, 38, 0.05);"> <th style="padding: 10px; border-bottom: 2px solid #dc2626; text-align: left; color: #dc2626; font-size: 14px; width: 30%;">Module / Component</th> <th style="padding: 10px; border-bottom: 2px solid #dc2626; text-align: left; font-size: 14px;">Issue & Troubleshooting Diagnostics</th> </tr> </thead> <tbody> <tr>The model card is truncated. Read the rest on Hugging Face.
- Task
- Image-Text-to-Text
- Library
- transformers
- Type
- qwen3_5
- License
- apache-2.0
- Languages
- en, zh, es, ru, ja
- Created
- May 28, 2026
- Updated
- May 28, 2026