Forcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDP
cs.CL, cs.LG
Submitted: 2026-11-05
Updated: 2026-11-05
Code: https://github.com/Saepi/WMDP-Safety-Interp
Terminology
Sources
- Eliciting Latent Predictions from Transformers with the Tuned Lens
- When Choices Become Risks: Safety Failures of Large Language Models under Multiple-Choice Constraints
- Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structure
- Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety
- An Adversarial Perspective on Machine Unlearning for AI Safety
- SG-Bench: Evaluating LLM Safety Generalization Across Diverse Tasks and Prompt Types
- SCORE: Systematic COnsistency and Robustness Evaluation for Large Language Models
- LLMs Encode Harmfulness and Refusal Separately
- Universal and Transferable Adversarial Attacks on Aligned 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