{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreia6hy6frgvvw6aukeosr5q7pnk5kkdovcgllfl5ju3qmkm347bp3a",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mmbzdcsh5sn2"
  },
  "path": "/t/looking-for-arxiv-cs-cr-endorsement/176132#post_1",
  "publishedAt": "2026-05-20T11:43:47.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "Log in to arXiv | arXiv e-print repository"
  ],
  "textContent": "Hi everyone,\n\nI’m submitting a research paper to arXiv (cs.CR / cs.AI) on real-time safety auditing of LLM agent workflows and am looking for an arXiv endorser.\n\nThe paper proposes a formal gating framework \" the _Agent-Trace-Verifier (ATV)_ \" that intercepts agent actions at runtime, compares them against a declarative task specification (CS-DAG), and flags semantic violations before execution. It introduces a risk-weighted gating function combining scope, irreversibility, and semantic novelty, with empirical results across 500 agent traces in 5 domains.\n\nIf anyone here has arXiv endorsement privileges and would be willing to help, I’d really appreciate it. Happy to share the abstract or full paper.\n\nThe endorsement process is very quick:\n\n  1. Click this link: Log in to arXiv | arXiv e-print repository\n  2. Or visit Log in to arXiv | arXiv e-print repository and enter code: _HWZDVO_\n  3. Confirm endorsement (takes ~30 seconds)\n\n\n\nWhy This Matters:\n\nAs LLM agents are deployed in high-stakes settings — healthcare, legal, finance — the gap between what an agent is _permitted_ to do (RBAC) and what it _should_ do (task intent) becomes a critical attack surface. ATV closes that gap with a lightweight semantic layer that works across any agent framework.\n\nThank you for considering this request. Happy to answer questions or discuss connections to your research.",
  "title": "Looking for arXiv cs.CR endorsement"
}