This is a heavily interactive web application, and JavaScript is required. Simple HTML interfaces are possible, but that is not what this is.
Post
Ekdeep Singh @ ICML
ekdeepl.bsky.social
did:plc:uecuvnzs25ci3mlbu6lwgyaq
To test the above claim, we compute the effect of shuffling a sequence on next-token probs: this breaks bigram stats, but preserves unigrams. We check how “retrieval-like” or memorization-based model behavior is by comparing predicted transitions’ KL to a random set of chains.
2025-02-16T18:57:39.611Z