A parameter-free sampling strategy called CUTS combined with Mixed-CUTS training prevents mode collapse in RL for saturated LLM reasoning tasks and raises AIME25 Pass@1 accuracy by up to 15.1% over standard GRPO.
Nature , volume=
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3representative citing papers
Active information seeking via search tools, when combined with multi-candidate context pruning during training, produces consistent gains on translation, health, and reasoning tasks over naive tool addition or no-tool baselines.
citing papers explorer
-
Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data
A parameter-free sampling strategy called CUTS combined with Mixed-CUTS training prevents mode collapse in RL for saturated LLM reasoning tasks and raises AIME25 Pass@1 accuracy by up to 15.1% over standard GRPO.
-
Context Training with Active Information Seeking
Active information seeking via search tools, when combined with multi-candidate context pruning during training, produces consistent gains on translation, health, and reasoning tasks over naive tool addition or no-tool baselines.
- Voice "Cloning" is Style Transfer