Collection by
1 day ago
1 day ago
-
Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning
Paper β’ 2506.01939 β’ Published β’ 127 -
Unsupervised Word-level Quality Estimation for Machine Translation Through the Lens of Annotators (Dis)agreement
Paper β’ 2505.23183 β’ Published β’ 2 -
Improved Representation Steering for Language Models
Paper β’ 2505.20809 β’ Published β’ 1 -
SAEs Are Good for Steering -- If You Select the Right Features
Paper β’ 2505.20063 β’ Published β’ 1
