SAEScientist-Bench: can AI agents autonomously run SAE interpretability research?

CASIA · hf · 2026-09-10

CASIA introduces SAEScientist-Bench, a benchmark testing whether AI agents can autonomously conduct sparse autoencoder (SAE) interpretability research.

The benchmark offers a targeted yardstick for agentic AI research capabilities in interpretability.

Original post →

More from coding & agent

coding & agent channel →