← Back to all news
arXiv AI By Nguyen Minh Thang, Trinh Thi Phuong

Rules, Traces and Routing Statistics: Training and Calibration Data for Open Models on ARC-AGI-2

Read the original on arXiv AI →

The paper reports on training and calibration experiments for three open models on the ARC‑AGI‑2 benchmark, using offline runs on four L4 GPUs for up to 12 hours. For the NVARC Qwen3‑4B grid model, a curated corpus of 12,000 executable rules and 95,993 tasks was built, and fine‑tuning on 1,920 real task records raised the semi‑private score from 28.96 to 33.06, with no synthetic‑task checkpoint outperforming the real‑task model. For the Qwen3.8‑27B reasoning model, fine‑tuning on 1,036 of its own reasoning traces and a specialized inference procedure increased the score from 26.81 to 44.03, while the Qwen3.8‑Flash‑Next mixture‑of‑experts model achieved a best score of 56.67 after expert pruning, routing statistics adaptation, and an FP8 indexer cache.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv AI
Aug 19

Beyond the Trace: Coupling an Interpretable Reasoning-State Readout to Native MoE Routing

The paper introduces a two‑level readout for mixture‑of‑experts reasoning models. First, it compresses the model’s internal reasoning states into a 64‑dimensional semantic frame (J64) that reveals process dynamics beyond the emitted trace. Second, it reconstructs this frame from native expert‑routing statistics (R64), achieving high correlation and preserving most predictive gains while enabling low‑overhead, test‑time decision making.

By Kang Chen, Sihan Zhao, Yixin Cao, Yugang Jiang
llms
More like this →
arXiv Computation and Language
2d ago

Concise-9B: Removing the Slack from a Reasoning Model's Thinking

The paper introduces Concise-9B, a version of the Qwen3.5-9B reasoning model that reduces unnecessary thinking tokens through concise self‑training. By applying two rounds of on‑policy rejection‑sampling fine‑tuning and a small share of answer‑synthesis targets, Concise-9B cuts token usage to roughly two‑thirds of the original model while maintaining comparable accuracy on a wide range of hard mathematics and reasoning benchmarks. The study also shows that most of the token savings occur after the answer has already formed, and that the approach’s effectiveness depends on the verbosity of the base model.

By Alexander Boesgaard Lorup
fine-tuningbenchmarks
More like this →
Hugging Face Trending Papers
Aug 18

Beyond the Trace: Coupling an Interpretable Reasoning-State Readout to Native MoE Routing

The paper introduces a two‑level internal readout for mixture‑of‑experts reasoning models. First, it compresses the model’s reasoning states into a 64‑axis semantic frame (J64) that reveals process states not captured by the emitted trace, improving held‑out AUC by 0.096–0.135. Second, it reconstructs this frame from native expert‑routing statistics (R64), achieving high correlation with J64 and preserving most of its predictive gain while enabling low‑overhead, test‑time decision making and improved routing policies.

llms
More like this →