Hanlin Zhang

hlzhang109.bsky.social

18 March 2026 19:53

Learning from feedback is instrumental but human preference data can be expensive. How much reward supervision could we get from raw web text instead, without human labels? Our work, a year of incredible effort by @fjxdaisy.bsky.social, advances pure RLHF training across multiple models and tasks.

Alexander Shakeel Bates

asbates.bsky.social

02 August 2025 17:18

Public access to the first fly connectome that spans the whole CNS - BANC!: codex.flywire.ai?dataset=banc Different from prior connectomes - it is brain + cord (think spinal cord) We use it to ‘embody’ the system and find it resembles ‘subsumption architecture’ doi.org/10.1101/2025...