9 papers
Post-Hoc Reasoning in Chain of Thought: Decoding and Steering Pre-Committed Answers
Kyle Cox, Darius Kianersi, Adrià Garriga-Alonso
As chain of thought (CoT) has become central to scaling reasoning capabilities in large language models (LLMs), it has also emerged as a promising tool for interpretability, sugges…
SynthSAEBench: Evaluating Sparse Autoencoders on Scalable Realistic Synthetic Data
David Chanin, Adrià Garriga-Alonso
Improving Sparse Autoencoders (SAEs) requires benchmarks that can precisely validate architectural innovations. Current LLM-based SAE benchmarks are too noisy to differentiate arch…
DiFR: Inference Verification Despite Nondeterminism
Adam Karvonen, Daniel Reuter, Roy Rinberg +3
As demand for LLM inference grows, it is becoming increasingly important that providers and their customers can verify that inference processes are performed correctly, without err…
Path Channels and Plan Extension Kernels: a Mechanistic Description of Planning in a Sokoban RNN
Mohammad Taufeeque, Aaron David Tucker, Adam Gleave +1
We partially reverse-engineer a convolutional recurrent neural network (RNN) trained with model-free reinforcement learning to play the box-pushing game Sokoban. We find that the R…
Interpreting Emergent Planning in Model-Free Reinforcement Learning
Thomas Bush, Stephen Chung, Usman Anwar +2
We present the first mechanistic evidence that model-free reinforcement learning agents can learn to plan. This is achieved by applying a methodology based on concept-based interpr…
Among Us: A Sandbox for Measuring and Detecting Agentic Deception
Satvik Golechha, Adrià Garriga-Alonso
Prior studies on deception in language-based AI agents typically assess whether the agent produces a false statement about a topic, or makes a binary choice prompted by a goal, rat…
Hypothesis Testing the Circuit Hypothesis in LLMs
Claudia Shi, Nicolas Beltran-Velez, Achille Nazaret +5
Large language models (LLMs) demonstrate surprising capabilities, but we do not understand how they are implemented. One hypothesis suggests that these capabilities are primarily e…
Adversarial Circuit Evaluation
Niels uit de Bos, Adrià Garriga-Alonso
Circuits are supposed to accurately describe how a neural network performs a specific task, but do they really? We evaluate three circuits found in the literature (IOI, greater-tha…
Planning in a recurrent neural network that plays Sokoban
Mohammad Taufeeque, Philip Quirke, Maximilian Li +4
Planning is essential for solving complex tasks, yet the internal mechanisms underlying planning in neural networks remain poorly understood. Building on prior work, we analyze a r…