RL Squeezes, SFT Expands: A Comparative Study of Reasoning LLMs

Documenti analoghi

• Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
  di: Matsutani, Kohsei, et al.
  Pubblicazione: (2026)
• Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
  di: Minegishi, Gouki, et al.
  Pubblicazione: (2025)
• Understanding Emergent Misalignment via Feature Superposition Geometry
  di: Minegishi, Gouki, et al.
  Pubblicazione: (2026)
• Safe Transformer: An Explicit Safety Bit For Interpretable And Controllable Alignment
  di: Feng, Jingyuan, et al.
  Pubblicazione: (2026)
• Towards Empirical Interpretation of Internal Circuits and Properties in Grokked Transformers on Modular Polynomials
  di: Furuta, Hiroki, et al.
  Pubblicazione: (2024)