{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreigsez5oufvtzt3ft3pa4lswrdagzpeoxdwin22y66gq7ulkap77c4",
"uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mjjldcmu5w32"
},
"path": "/t/hidden-state-signals-in-iterative-llm-repair-what-cosine-similarity-at-layer-27-actually-tells-you/174143#post_3",
"publishedAt": "2026-04-15T08:12:15.000Z",
"site": "https://discuss.huggingface.co",
"textContent": "Just bumping this — still happy to discuss the methodology, the hidden-state extraction setup, or the monotonic controller design. Would be especially interested to hear from anyone who’s experimented with similar inference-time signals on other models or tasks.",
"title": "Hidden-State Signals in Iterative LLM Repair: What Cosine Similarity at Layer 27 Actually Tells You"
}