Downloads · 30 days
9
21% of all-time downloads
satoyutaka/LLM2025_main_0_DPO11
LLM2025_main_0_DPO11 is a text generation model from satoyutaka. Use it when you need the model to write or continue text. It is set up for peft. The card lists the license as apache-2.0.
This repository provides a LoRA adapter for Qwen/Qwen3-4B-Instruct-2507. It represents the "Surgical Strike" version (v11) of the "struct-eval-comp" series, specifically designed to eliminate Markdown formatting bias…
Downloads · 30 days
9
21% of all-time downloads
All-time downloads
42
Public
Repo size
96.4 MB
Likes
0
Public
Click a slice to open those files.
.safetensors85 MB · 84%
From the Hugging Face model README
This repository provides a LoRA adapter for Qwen/Qwen3-4B-Instruct-2507. It represents the "Surgical Strike" version (v11) of the "struct-eval-comp" series, specifically designed to eliminate Markdown formatting bias while preserving the core intelligence of the Attention layers.
Based on the failure of v8 (structural collapse) and the stagnation of v9/v10, v11 adopts a "Partial Freezing" philosophy. By using an ultra-low learning rate and a specific Beta value, we effectively protect the reasoning logic stored in the Attention modules while fine-tuning the MLP layers for raw output excellence.
The primary goal of v11 is "Intelligence Preservation through Attention Stability." Previous iterations showed that aggressive DPO across all layers could degrade the model's ability to handle complex JSON nesting. V11 resolves this by applying a micro-learning rate (5e-7), which ensures that the Attention weights—responsible for the model's 0.77136 logic score—remain virtually unchanged, while the MLP layers are nudged to drop the persistent ```json backticks.
from unsloth import FastLanguageModel
import torch
base = "Qwen/Qwen3-4B-Instruct-2507"
adapter = "satoyutaka/LLM2025_main_0_DPO11"
model, tokenizer = FastLanguageModel.from_pretrained(
model_name = base,
max_seq_length = 2048,
load_in_4bit = True,
)
model.load_adapter(adapter)
このリポジトリは、Qwen/Qwen3-4B-Instruct-2507 用のLoRAアダプターです。 struct-eval-comp プロジェクトの「精密狙撃(V11)」モデルであり、V7/V9で確立した高い推論知能を物理的に保護しつつ、フォーマットの矯正のみを完遂することを目指しています。
V11の最大の特徴は、**「推論の核となるAttention層のロジックを実質的に固定し、出力スタイルを司るMLP層のみを微調整する」**というアプローチにあります。
V10までの分析により、全層(q, k, v, o等を含む)に対して強いDPO圧力をかけると、構造化能力そのものが毀損される(V8のケース)ことが判明しました。 V11では、超低学習率(5e-7)とターゲットモジュールの厳選、そしてBeta値の最適化を組み合わせることで、Attention層に蓄積された「0.77水準の知能」を損なうことなく、Markdownバイアスのみをピンポイントで中和しています。
(英語セクションの Usage コードを参照してください)