Parts-of-Speech as Emergent Categories in SAE Latent Space
The Linear Representation Hypothesis Needs a Group Action
Deep Noir: Autonomous Steering Discovery via Architectural Chronometry in Transformer Models
Large Language Models Develop Belief State Geometry In-Context
1
The Router Within: Eliciting Native Skill Routing from a Frozen LLM
1
Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration
From Parameters to Answers: How LLMs Retrieve and Use Their Internal Knowledge
1
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
2
LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders
1
Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs
1