The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models

Documenti analoghi

• Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training
  di: Wang, Mengru, et al.
  Pubblicazione: (2025)
• Dancing with Critiques: Enhancing LLM Reasoning with Stepwise Natural Language Self-Critique
  di: Li, Yansi, et al.
  Pubblicazione: (2025)
• Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs
  di: Wang, Yue, et al.
  Pubblicazione: (2025)
• MSL: Not All Tokens Are What You Need for Tuning LLM as a Recommender
  di: Wang, Bohao, et al.
  Pubblicazione: (2025)
• Trust, But Verify: A Self-Verification Approach to Reinforcement Learning with Verifiable Rewards
  di: Liu, Xiaoyuan, et al.
  Pubblicazione: (2025)