
Henry Freed
@hbfreed
pawpaw/acc
Tots ‘n’ goop
ID: 1796570417051627520
31-05-2024 15:53:35
75 Tweet
26 Followers
152 Following



The EleutherAI interpretability team is releasing a new open source pipeline for automatically interpreting SAE features and neurons in LLMs, using LLMs. We also introduce five new, efficient techniques for evaluating the quality of explanations. arxiv.org/abs/2410.13928







Very impressed by Voyage AI by MongoDB's rerankers so far! rerank-2-lite basically on par with cohere 3.5 in my testing. Substantially cheaper too: ~16 cents vs $2.58 for cohere 3.5 for the same suite.




