{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreifissx6vyfxhm6q3brxcj4k4y6kwghscdi6vhyctjs7hkzfjujc5y",
"uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mgnviuey2px2"
},
"path": "/t/looking-for-endorsor-for-arxiv-submission-cs-lg/168300#post_14",
"publishedAt": "2026-03-09T21:48:05.000Z",
"site": "https://discuss.huggingface.co",
"tags": [
"storm-safety-proofs/specs/SYSTEM_SPECIFICATION.pdf at main · kdschulte/storm-safety-proofs · GitHub",
"GitHub - kdschulte/storm-safety-proofs: Mathematical framework for proving verifiable safety properties (attribution, guaranteed forgetting, compositional non-degradation, memory isolation) in modular compositional AI systems. · GitHub"
],
"textContent": "Hi, I’m in a similar position and would really appreciate any help.\n\nI’m an independent researcher based in London looking for a cs.LG endorsement. My paper looks at whether modular AI systems, where frozen expert modules are composed through sparse routing, can make safety properties like guaranteed forgetting and non-degradation formally provable. It’s framed as open problems inviting the formal methods and learning theory communities to contribute.\n\nPaper: storm-safety-proofs/specs/SYSTEM_SPECIFICATION.pdf at main · kdschulte/storm-safety-proofs · GitHub\n\nRepo: GitHub - kdschulte/storm-safety-proofs: Mathematical framework for proving verifiable safety properties (attribution, guaranteed forgetting, compositional non-degradation, memory isolation) in modular compositional AI systems. · GitHub\n\nEndorsement code: ZYHFXN\n\nHappy to answer any questions about the work.\n\nThanks,\nKim",
"title": "Looking for endorsor for arXiv Submission (cs.LG)"
}