Self-supervised confidence fine-tuning makes reasoning more efficient, reducing generated tokens by up to 25\% at matched accuracy across Gemma, Qwen, Nemotron, and GPT-OSS models on mathematical, scientific, and coding reasoning benchmarks, with efficiency gains comparable to methods that explicitly optimize for short...
Parsa Hosseini, Akasha Tigalappanavara, Sumit Nawathe et al.· 0 citations
An exam-style evaluation framework is introduced for studying the global budget allocation of reasoning language models when multiple problems share an end-to-end cost or latency constraint, in which a model must distribute one shared token budget across questions with different difficulty and point values.
Chenrui Fan, Yize Cheng, Ming Li et al.· 0 citations
A two-phase continual-learning evaluation built from hard tasks in Terminal-Bench 2.0 found that optimization gains compounded only when regression control was built into the optimization loop, providing an inductive bias against shortcut solutions that fail to generalize.
Wenxiao Wang, Priyatham Kattakinda, S. Feizi· arXiv.org· 2 citations
The results suggest that current training-based methods do not yet extract insights beyond their training data, and motivate benchmarks that explicitly require interpretability methods to demonstrate such insights.
S. Balasubramanian, S. Feizi· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
