Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
cs.LG, cs.AI, cs.CL
Submitted: 2026-02-11
Updated: 2026-11-12
Terminology
Sources
- Improving Steering Vectors by Targeting Sparse Autoencoder Features
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders
- Language Models are Injective and Hence Invertible
- Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders
- Mixture-of-Depths: Dynamically allocating compute in transformer-based language models
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Gemma 2: Improving Open Language Models at a Practical Size
- The Llama 3 Herd of Models
- Steering Language Models With Activation Engineering
Related papers
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis
- (Mis)generalization of Helpful-Only Fine-tuning