{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:N3JO3GFRPGDLLA7BG2HTKZU74V","short_pith_number":"pith:N3JO3GFR","schema_version":"1.0","canonical_sha256":"6ed2ed98b17986b583e1368f35669fe56e9f6e41e5b88856dd50473350093d5d","source":{"kind":"arxiv","id":"2602.20042","version":2},"attestation_state":"computed","paper":{"title":"AI Alignment Breaks at the Edge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Carl Yang, Han Bao, Xiangliang Zhang, Xiaoda Wang, Yanfang Ye, Yue Huang, Yujun Zhou, Zheyuan Zhang","submitted_at":"2026-02-23T16:51:43Z","abstract_excerpt":"General Alignment has improved average-case helpfulness and safety, but current alignment practice still rewards confident, single-turn responses. The problem is not only that models fail on edge cases; it is that current evaluation makes many of these failures hard to see. We take the position that alignment must move beyond average-case evaluation by making failures under value conflict, plural stakeholder disagreement, and epistemic ambiguity visible and actionable. Scalar rewards compress diverse values into a single number; data and evaluation regimes collapse, filter, or fail to elicit t"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2602.20042","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-02-23T16:51:43Z","cross_cats_sorted":[],"title_canon_sha256":"8a63c19d63df867d7ce2d5ec465c41cf611e31d09c4998b81b5f31cc139a8c76","abstract_canon_sha256":"3538c92597bd73f02f2c87aecc3bfc4b21ea52a4f657e8b48b397ca4ee29aae3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:26.372888Z","signature_b64":"AgMscxBVTeP9hKKMjiFiohpi1DOQNN7xDFI1yl5cyFZQGepvGNznYOOuiRTBDtMAxQuyZUKY6budr1dBHFy8CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6ed2ed98b17986b583e1368f35669fe56e9f6e41e5b88856dd50473350093d5d","last_reissued_at":"2026-05-20T00:04:26.372054Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:26.372054Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AI Alignment Breaks at the Edge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Carl Yang, Han Bao, Xiangliang Zhang, Xiaoda Wang, Yanfang Ye, Yue Huang, Yujun Zhou, Zheyuan Zhang","submitted_at":"2026-02-23T16:51:43Z","abstract_excerpt":"General Alignment has improved average-case helpfulness and safety, but current alignment practice still rewards confident, single-turn responses. The problem is not only that models fail on edge cases; it is that current evaluation makes many of these failures hard to see. We take the position that alignment must move beyond average-case evaluation by making failures under value conflict, plural stakeholder disagreement, and epistemic ambiguity visible and actionable. Scalar rewards compress diverse values into a single number; data and evaluation regimes collapse, filter, or fail to elicit t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.20042","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.20042/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2602.20042","created_at":"2026-05-20T00:04:26.372193+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.20042v2","created_at":"2026-05-20T00:04:26.372193+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.20042","created_at":"2026-05-20T00:04:26.372193+00:00"},{"alias_kind":"pith_short_12","alias_value":"N3JO3GFRPGDL","created_at":"2026-05-20T00:04:26.372193+00:00"},{"alias_kind":"pith_short_16","alias_value":"N3JO3GFRPGDLLA7B","created_at":"2026-05-20T00:04:26.372193+00:00"},{"alias_kind":"pith_short_8","alias_value":"N3JO3GFR","created_at":"2026-05-20T00:04:26.372193+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.07655","citing_title":"Guardian-as-an-Advisor: Advancing Next-Generation Guardian Models for Trustworthy LLMs","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V","json":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V.json","graph_json":"https://pith.science/api/pith-number/N3JO3GFRPGDLLA7BG2HTKZU74V/graph.json","events_json":"https://pith.science/api/pith-number/N3JO3GFRPGDLLA7BG2HTKZU74V/events.json","paper":"https://pith.science/paper/N3JO3GFR"},"agent_actions":{"view_html":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V","download_json":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V.json","view_paper":"https://pith.science/paper/N3JO3GFR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.20042&json=true","fetch_graph":"https://pith.science/api/pith-number/N3JO3GFRPGDLLA7BG2HTKZU74V/graph.json","fetch_events":"https://pith.science/api/pith-number/N3JO3GFRPGDLLA7BG2HTKZU74V/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V/action/storage_attestation","attest_author":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V/action/author_attestation","sign_citation":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V/action/citation_signature","submit_replication":"https://pith.science/pith/N3JO3GFRPGDLLA7BG2HTKZU74V/action/replication_record"}},"created_at":"2026-05-20T00:04:26.372193+00:00","updated_at":"2026-05-20T00:04:26.372193+00:00"}