{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreifq7y4othrefirzcca372xwuv2qhvumfrufgofo33uk2ismavewmy",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mo2h4qwaisg2"
  },
  "path": "/t/can-an-llm-lose-conceptual-continuity-while-remaining-coherent/176469#post_12",
  "publishedAt": "2026-06-11T22:30:46.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "Thanks, that makes sense.\n\nYes, I am actually working on the epistemic layer. My focus is not mainly cache optimization, but preserving epistemic objects outside the model context: claims, constraints, evidence handles, unresolved conflicts, role state, and method decisions.\n\nSo I see your TIS/KV work and my layer as complementary:\n\n  * your layer decides what should remain mechanically salient during inference\n  * the epistemic layer decides what must remain valid, classified, and reusable across inference steps\n\n\n\nFor LITM, this distinction may matter a lot. The problem is not only whether an item is still accessible, but whether the system still knows why it matters and in which epistemic status it may be reused.\n\nSo yes, if you share your NIAH or LITM results later, I would be interested in testing how such a cache/salience layer could interact with an external epistemic state layer.",
  "title": "Can an LLM lose conceptual continuity while remaining coherent?"
}