Open-weight models do not report injected pain states
Across 52,920 reports, excess pain claims stayed under a few points, while probes still recovered the steered state.
Stories per day · 30 days
2 RSS
Relatedalignment2qwen2abliteration1aime1harmbench1hmmt1internvl1jailbreak1llm-refusal1llm-supply-chain1mllm1model-introspection1multimodal-safety1open-weight-models1reasoning-models1recommendation-hijack1refusal1refusal-steering1
15 stories, newest first · one card per story
Stories tagged #activation-steering
Paper finds trojan risk in LLM steering-vector plugins
Authors say plugin IDH reached up to 97.8% attack success on five LLMs while keeping high refusal rates.
Steered-teacher self-distillation cuts reference hallucinations in thinking models
Activation steering is applied only to the teacher; the student is not steered, and AIME and HMMT accuracy rises over vanilla OPSD.
VLM safety steering differs by layer depth, study finds
On Qwen3-VL-4B, early vision-layer steering moved unsafe rejection between 6% and 96% without retraining.
WRDO learns weighted directions to change LLM refusal behavior
The method is tested on three open-weight models with HarmBench and can also be reversed to restore refusal.
MARS steers multimodal safety from textual refusal directions
The training-free method was tested on five state-of-the-art MLLMs for safety, utility, and video jailbreaks.
Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation Steering
A closed-loop steering attack injects targeted demographic bias into diffusion language models during denoising.
Certified Mechanistic Edits: Behavioral Guarantees for Skill Removal and Preservation
Researchers certify that mechanistic edits provably remove one neural network skill while preserving another, replacing test-only validation with formal behavioral guarantees.
SteerProbe: Learning to Bypass Safety Steering in Vision-Language Models
SteerProbe bypasses vision-language model safety steering, lifting average harmful responses from 7.43% to 18.36%.
A mechanistic study of language model introspection
Researchers identify attention-head circuits that let language models detect and localize hidden activation interventions.
Persona Dosing: Calibrated Activation Steering for Graded Trait Control
PersonaDose calibrates activation steering so models express requested persona intensities.
Learning to Steer, Steering to See: Unveiling the Geometry of RLVR in Large Language Models via Trainable Vectors
Vector steering finds a low-dimensional manifold behind RLVR gains and a stabilizer that preserves them.
When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs
SAKIKO finds activation steering can raise tool-use scores while corrupting many previously correct LLM decisions.
Deep Noir: Autonomous Steering Discovery via Architectural Chronometry in Transformer Models
Deep Noir automates activation steering parameter discovery and finds steering creates a prompt-injection attack surface that widens with steering magnitude.
Steering Geometry: Validating Human Value Geometry in LLM Steering Space
Researchers show LLM activation steering vectors encode human value geometry predicted by Schwartz's theory, with fidelity improving at scale but degrading after instruction tuning.