Why Does Reinforcement Learning Generalize? A Feature-Level Mechanistic Study of Post-Training in Large Language Models

Documenti analoghi

• The Lookahead Limitation: Why Multi-Operand Addition is Hard for LLMs
  di: Baeumel, Tanja, et al.
  Pubblicazione: (2025)
• Sparse Subnetwork Enhancement for Underrepresented Languages in Large Language Models
  di: Gurgurov, Daniil, et al.
  Pubblicazione: (2025)
• Probing Context Localization of Polysemous Words in Pre-trained Language Model Sub-Layers
  di: Vijayakumar, Soniya, et al.
  Pubblicazione: (2024)
• Disentangling Mathematical Reasoning in LLMs: A Methodological Investigation of Internal Mechanisms
  di: Baeumel, Tanja, et al.
  Pubblicazione: (2026)
• On Multilingual Encoder Language Model Compression for Low-Resource Languages
  di: Gurgurov, Daniil, et al.
  Pubblicazione: (2025)