VERS: Steering Reasoning Models by Editing Their Activation Explanations
cs.CL, cs.LG
Submitted: 2026-11-02
Updated: 2026-11-02
Code: https://github.com/Strongich/doomloop-interp
Terminology
Sources
- Can Aha Moments Be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thought
- Steering Llama 2 via Contrastive Activation Addition
- Understanding Reasoning in Thinking Language Models via Steering Vectors
- Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
- Steering Language Models With Activation Engineering
- Representation Engineering: A Top-Down Approach to AI Transparency
- Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs
- s1: Simple test-time scaling
- Mitigating Overthinking in Large Reasoning Models via Manifold Steering
- Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency
- Dynamic Early Exit in Reasoning Models
- Thinking Past the Answer: Evaluating Harmful Overthinking in Large Reasoning Models
- Transferring Linear Features Across Language Models With Model Stitching
- Activation Space Interventions Can Be Transferred Between Large Language Models
- Linear Representation Transferability Hypothesis: Leveraging Small Models to Steer Large Models
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Qwen3 Technical Report
- Measuring Mathematical Problem Solving With the MATH Dataset
- Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions
- LoRA: Low-Rank Adaptation of Large Language Models
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data