Viewing profile — mlin4589

The reality, I suspect is that internally models are likely modeling these alignment features such as refusals as a secondary filter. In fact, for many models you can remove refusa…

Calibration (in a binary context) basically means that the confidence of a model/score matches the probability that a particular label is positive or not. For instance, a calibrate…

Good question! We do know from OpenAI's system card from GPT-4 that the post-trained RLHF model is significantly less calibrated compared to the pre-trained model, so it's a matter…

We're building real-time abuse prevention systems. Hiring Full-stack, frontend, and MLEs across product and PoCs. https://withintrinsic.com/careers