DPO Debate: Is RL needed for RLHF?

All things as we cannot settle if DPO or RL is better. At least it is a good exercise.
Slides: docs.google.com/presentation/...
Recent blog post of mine on DPO: www.interconnects.ai/p/rlhf-p...
DPO Paper: arxiv.org/abs/2305.18290

[@pruff3]

Finally, I love DP....oh

[@Kopern1k990]

Thanks for sharing your thoughts

[@spartaleonidas540]

Will LMsys release their chatbot arena preference dataset

[@SantoshGupta-jn1wn]

great video, thanks

[@MacProUser99876]

How DPO works under the hood: th-cam.com/video/Ju-pFJNfOfY/w-d-xo.html

[@aojing]

@5:00 (14)

[@mohamedfouad1309]

❤