Dipika Khullar

If you take the weight difference between a reasoning model and its non-reasoning instruct counterpart, and then apply more of that difference to the reasoning model, you get what we call an overthinking model. Overthinking models are usually worse at keeping secrets. This is good, because models should (generally) be...

Paper A common pattern in AI evaluation pipelines involves using an LLM to generate an action, then using the same model to evaluate whether that action is safe or correct. This appears in coding agents that review their own pull requests, tool-using assistants that assess whether their proposed actions are...