Skip to content
conftrace_

critical token

2 papers

Explore in graph

Co-occurring keywords · papers mentioning both

kl penalty (1) direct preference optimization (dpo) (1) gsm8k (1) llm reasoning capability (1) rollout sampling (1) token-level contrastive estimation (1) large language model (1) reinforcement learning fine-tuning (1) token-level exploration (1) reward signal (1)

Papers

Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM’s Reasoning Capability ICML 2025 Ignore the KL Penalty! Boosting Exploration on Critical Tokens to Enhance RL Fine-Tuning NAACL 2025