CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features

Documenti analoghi

• Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
  di: Cho, Seonglae, et al.
  Pubblicazione: (2026)
• Curveball Steering: The Right Direction To Steer Isn't Always Linear
  di: Raval, Shivam, et al.
  Pubblicazione: (2026)
• Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
  di: Patel, Het, et al.
  Pubblicazione: (2026)
• Entropy-Based Measurement of Value Drift and Alignment Work in Large Language Models
  di: Fadli, Samih
  Pubblicazione: (2025)
• Painless Activation Steering: An Automated, Lightweight Approach for Post-Training Large Language Models
  di: Cui, Sasha, et al.
  Pubblicazione: (2025)