AI agents show promise but struggle to understand model features fully
SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Understanding what AI models learn is important to make them safer and more reliable. The authors created a test called SAEScientist-Bench to see if AI agents can act like scientists by using special tools called sparse autoencoders to study model features on their own. Their experiments show that while AI agents can find important features and rule out wrong choices, they often misunderstand some key measurements compared to human experts. This work helps measure how well AI can independently understand other AI, which is useful for future automated research.
Recursive self-improvementSparse Autoencoders (SAEs)Mechanistic interpretabilityContrastive probesFeature discoveryActivation rankConcept selectivityCausal steeringAI alignmentAutonomous AI research
Authors
Yuqiao Tan, Shizhu He, Jun Zhao, Kang Liu
Abstract
While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment. Mechanistic interpretability tools are essential to bridge this gap, among which Sparse Autoencoders (SAEs) serve as a cornerstone by isolating interpretable features for model inspection and steering. In this paper, we introduce SAEScientist-Bench to evaluate whether AI agents can act as scientists utilizing SAE tools for autonomous mechanistic discovery. Given a target concept, an agent designs contrastive probes and navigates a Gemma Scope dictionary of 131K+ features in Gemma-2-9B-IT to discover the optimal feature, evaluated against curated expert reference features anchored on Neuronpedia across activation rank, concept selectivity on contrastive texts, and causal steering. Across 10 agent configurations and 20 tasks, frontier agents demonstrate genuine discovery capabilities and lead different evaluation dimensions, but remain well behind the expert baseline, approaching expert levels on separating target concepts from contrastive controls while lagging substantially in causal generation steering. Further analysis reveals that although agents can design contrasts to rule out spurious candidates, they frequently misinterpret experimental measurements. These results establish experimental model understanding as a measurable capability for closed-loop autonomous AI R&D. Our code is available at https://github.com/Trae1ounG/SAEScientist.