Model Details

This repo has multimodal SAE's trained on Qwen 2.5 hidden states.

Uses

These SAE's are intended to be used for downstream steering tasks, specificallly for multimodla environments like browser use.

Develop automated intrepretabillity pipeline to label human readable features for the SAE's. Work on automated steering browsing agent.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support