Post
brendan chambers
societyoftrees.bsky.social
did:plc:caixreruxekqboehleym5bs4
Less discussed though , their choice of reverse-KL is also worth noting in particular.
- It frees the student to drop modes (since divergence = 0 where the student model has no coverage)
- It adds optimization pressure even where the teacher distribution has no coverage (e.g. sampling noise)
2025-10-28T18:37:39.593Z