Large Language Models Generate Harmful Responses Using a Distinct Mechanism, Shared Across Harm Types
Paper • 2604.09544 • Published • 7
AI & ML interests
Principled evaluation of mechanistic interpretability methods.
Recent Activity
View all activity
hadasor
authored 6
papers about 2 months ago
Agents of Chaos
Paper • 2602.20021 • Published • 36
Inside-Out: Hidden Factual Knowledge in LLMs
Paper • 2503.15299 • Published • 56
Position-aware Automatic Circuit Discovery
Paper • 2502.04577 • Published • 1
Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelines
Paper • 2403.05846 • Published • 1
Unified Concept Editing in Diffusion Models
Paper • 2308.14761 • Published • 2
hadasor
submitted a
paper to Daily Papers 6 months ago
nepp1d0
authored a
paper 7 months ago
amueller
updated a
Space 8 months ago
hij
authored a
paper 11 months ago
hij
authored 3
papers about 1 year ago
AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders
Paper • 2501.17148 • Published • 1
LLMs Encode Harmfulness and Refusal Separately
Paper • 2507.11878 • Published • 1
Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors
Paper • 2505.11770 • Published • 2
atticusg
authored a
paper over 1 year ago
belinkov
authored a
paper almost 2 years ago
hadasor
authored a
paper almost 2 years ago
belinkov
authored a
paper about 2 years ago
hadasor
authored a
paper about 2 years ago
atticusg
authored a
paper about 2 years ago
AdamBelfki
authored a
paper about 2 years ago