Sparse Autoencoders (SAEs) are popular, with 10+ new approaches proposed in the last year. How do we know if we are making progress? The field has relied on imperfect proxy metrics. We are releasing SAE Bench, a suite of 8 SAE evaluations! Project co-led with Can Rager 🧵

Excited to share our work on Matryoshka SAEs - a new variant of sparse autoencoders that learn features at multiple levels of abstraction by splitting the dictionary into nested groups of latents of increasing size! w/ Patrick Leask and Neel Nanda 🪆🪆🪆

Working with Neel as a MATS Research scholar has been a fantastic experience. I highly recommend applying!

Imagine if ChatGPT highlighted every word it wasn't sure about. We built a streaming hallucination detector that flags hallucinations in real-time.

I'm excited that, this year, interpretability finally works well enough to be practically useful in the real world! We found that, with enough effort into dataset construction, simple linear probes are cheap, real-time, token level hallucination detectors and beat baselines

🕳️🐇Into the Rabbit Hull – Part II Continuing our interpretation of DINOv2, the second part of our study concerns the geometry of concepts and the synthesis of our findings toward a new representational phenomenology: the Minkowski Representation Hypothesis