<h1 align="center"><strong>K2-Horizon-0.9B-Uno</strong></h1>
This model accompanies the paper Unlocking Lossless Speedups in LLMs via Discrete Diffusion.
Project page: https://s-sahoo.github.io/uno/
Code: https://github.com/ifm-ai/uno
K2-Horizon-0.9B-Uno is a conditional-LoRA adapter for diffusion-style decoding with K2-Horizon-0.9B.
This repository contains the adapter only. The base-model weights are hosted separately in IFM/K2-Horizon-0.9B.
Code
Selected evaluation scripts are available in scripts/k2_horizon. The full evaluation suite will be released soon.
Evaluation results
The main number is the benchmark accuracy and the subscript is TPF. --
denotes an unavailable TPF.
<!-- TABLE:START -->
<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;margin:0 auto;padding:8px 0 16px;overflow-x:auto"><table style="display:table;width:100%;table-layout:fixed;border-collapse:collapse;font-size:12px;margin:0"><thead><tr><th style="width:60%;border-bottom:none"></th><th style="padding:6px 4px 2px;text-align:center;font-size:11px;font-weight:600;letter-spacing:0.04em;text-transform:uppercase;opacity:0.65;border-bottom:1px solid rgba(128,128,128,0.25)">Ours</th></tr><tr><th style="padding:10px 6px;text-align:left;border-bottom:2px solid #2450D6">Benchmark</th><th style="padding:10px 3px;text-align:center;font-weight:600;border-bottom:2px solid #2450D6;color:#2450D6;background:rgba(36,80,214,0.08)"><strong>Uno</strong><br><sub>0.9B</sub></th></tr></thead><tbody><tr><td colspan="2" style="padding:6px 10px;font-weight:600;font-size:12.5px;color:#2450D6;border-bottom:1px solid rgba(36,80,214,0.25);background:rgba(36,80,214,0.12)">Long-Context Reasoning</td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">AA-LCR</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>18.0</strong><sub>1.81</sub></td></tr><tr><td colspan="2" style="padding:6px 10px;font-weight:600;font-size:12.5px;color:#2450D6;border-bottom:1px solid rgba(36,80,214,0.25);background:rgba(36,80,214,0.12)">Science and Knowledge</td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">ARC-Challenge</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>78.2</strong><sub>1.46</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">GPQA-Diamond (avg@16)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>27.3</strong><sub>1.59</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">HLE (Full)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>5.4</strong><sub>1.53</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">AA-Omniscience</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>7.2</strong><sub>1.64</sub></td></tr><tr><td colspan="2" style="padding:6px 10px;font-weight:600;font-size:12.5px;color:#2450D6;border-bottom:1px solid rgba(36,80,214,0.25);background:rgba(36,80,214,0.12)">Math</td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">AIME 2024</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>43.3</strong><sub>1.50</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">AIME 2025 (avg@16)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>41.7</strong><sub>1.54</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">AIME 2026 (avg@16)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>48.5</strong><sub>1.52</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">GSM8K</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>88.2</strong><sub>1.58</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">HMMT February 2026 (avg@16)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>25.8</strong><sub>--</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">MATH500 (Full)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>86.2</strong><sub>1.57</sub></td></tr><tr><td colspan="2" style="padding:6px 10px;font-weight:600;font-size:12.5px;color:#2450D6;border-bottom:1px solid rgba(36,80,214,0.25);background:rgba(36,80,214,0.12)">Coding</td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">HumanEval</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>62.8</strong><sub>1.79</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">HumanEval+ (pass@1)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>79.9</strong><sub>--</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">LiveCodeBench v6 (avg@3)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>37.4</strong><sub>--</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">MBPP</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>70.4</strong><sub>1.53</sub></td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">MBPP+ (pass@1)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>68.0</strong><sub>--</sub></td></tr><tr><td colspan="2" style="padding:6px 10px;font-weight:600;font-size:12.5px;color:#2450D6;border-bottom:1px solid rgba(36,80,214,0.25);background:rgba(36,80,214,0.12)">Instruction Following</td></tr><tr><td style="padding:6px 4px 6px 10px;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;font-weight:600;line-height:1.2">IFEval (strict instruction)</td><td style="padding:6px 3px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);vertical-align:middle;font-size:12.5px;line-height:1.2;background:rgba(36,80,214,0.08)"><strong>80.8%</strong><sub>1.72</sub></td></tr></tbody></table></div>
<!-- TABLE:END -->
Conversion provenance
conversion_summary.json records the adapter hash, source checkpoint,
base-weight match, and tensor-key compatibility validation.
The public K2-Horizon base is a schema migration of the exact local base used
for training. The base weight, index, and tokenizer hashes match; only the
public Python architecture name changed from K2Aurora to K2Horizon. All 392
adapter tensors map to valid target weights in the public base model.
Citation
If you find this model useful, please cite:
@misc{k2_horizon_09b_uno,
title = {K2-Horizon-0.9B-Uno},
author = {Institute of Foundation Models},
year = {2026},
howpublished = {\url{https://huggingface.co/IFM/K2-Horizon-0.9B-Uno}},
}
