Post

contrastive-lm.notion.site Interesting write up. No paper yet unfortunately. If some of the claims hold true it would be a pretty incredible leap. Their SOTA DeepSWE scores come from generating rollouts using Opus/Fable and then using their model to select the "right" action.

contrastive-lm.notion.siteContrastive Language Models | NotionA System One Model for Fast and Generalizable Decision-Making

Andrew Gross

@gross.systems

Does seem like they used a similar approach to CLIP but for non-image inputs? Need to spend more time understanding the training data vs what is used for a fine tune and why etc.

24/9/26, 12:21 pm0 reposts0 likes

No replies yet.