Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
Paper • 2602.10437 • Published • 2
🏗️ Building on HF
Seonglae Cho
seonglae
AI & ML interests
NLP, Humanoid, AGI
Recent Activity
authored a paper 4 days ago
OptimismBench: Forecasting Bias and the Alignment Effect in Language Model Judgment upvoted a paper 5 days ago
OptimismBench: Forecasting Bias and the Alignment Effect in Language Model JudgmentOrganizations
upvoted a paper 3 days ago
authored a
paper 4 days ago
upvoted a paper 5 days ago
authored 2
papers 7 days ago
upvoted a paper 7 days ago
updated a
Space 29 days ago
updated 7
Spaces about 1 month ago
Control Reinforcement Learning
🎛
Explore LLM token decisions with feature‑driven visualizations
PaaT: Probe as a Tool for Proprioceptive Language Agents
🔀
CorrSteer: Correlation-Based Steering of Language Models via Sparse Autoencoders
🧭
4
Steer language model output by clicking visual layers
OptimismBench
🎲
Confidence Manifold
🧭
AgentRoom
🏠
Watch two AI agents code together in a shared workspace
Agent State Graph — Dashboard
📊
published 5
Spaces about 1 month ago
updated a
dataset about 1 month ago