← Back to all news
arXiv Machine Learning By Chuqin Geng, Li Zhang, Haolin Ye, Mark Zhang, Luke Zhang, Xujie Si

Are We Recovering Mechanisms? Objective-Level Recovery Gaps in Mechanistic Interpretability

Read the original on arXiv Machine Learning →

The paper investigates mechanistic interpretability, focusing on how automated circuit discovery is evaluated. It shows that the commonly used faithfulness objective can favor circuits that reproduce a model’s behavior poorly, creating an objective-level recovery gap. Experiments on four human-reference tasks and InterpBench reveal that many discovery methods misrank candidate circuits, and that restoring excluded signals can correct most of these misrankings without altering the circuits’ behavior.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv AI
Sep 4

ObserverBench: Testing Mechanistic Estimates for Intervention and Control

ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, showing that accurate predictions do not always lead to better decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B demonstrate that observers trained on action loss can reduce deployment loss, while traditional metrics like AUROC may rank monitors differently from actual performance.

By Vijay Erramilli
llmsbenchmarkssafety
More like this →
Hugging Face Trending Papers
Sep 2

ObserverBench: Testing Mechanistic Estimates for Intervention and Control

ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, demonstrating that accurate average estimates can still lead to poor decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B show that observers trained on action loss tend to select lower‑loss actions, while traditional metrics like AUROC can rank monitors differently from deployment loss, highlighting the need for task‑specific evaluation.

llmsbenchmarkssafety
More like this →