Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
0 results for “mathematical reasoning”
Stochastic Teacher Intervention for Agentic On-Policy Distillation
Researchers propose STI-OPD, a new stochastic teacher intervention framework to improve on-policy distillation for multi-turn agentic language model training by dynamically replacing student actions with teacher actions based on policy discrepancy, thereby mitigating error accumulation and improving supervision reliability.
Oct 9, 2026
INSPIRE: An Internalize-Then-Improve Approach for Example-Driven Mathematical Reasoning
A new research paper introduces INSPIRE, a two-stage training method for LLMs that aims to improve example-driven mathematical reasoning by first internalizing the strategy and then refining correctness — addressing a gap in how models learn conceptual understanding versus pattern-matching.
Aug 31, 2026
Probing the Origins of Reasoning Performance: Representational Quality for Mathematical Problem-Solving in RL vs. SFT Fine-Tuned Models
A new arXiv preprint investigates why reinforcement learning (RL)-fine-tuned large language models outperform supervised fine-tuned (SFT) models on mathematical reasoning tasks, identifying representational differences in hidden-state structure and layer-wise importance as key mechanistic drivers.
Jul 31, 2026