{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XY2G3K54XRXGW4UQENJRW7XC2Y","short_pith_number":"pith:XY2G3K54","schema_version":"1.0","canonical_sha256":"be346dabbcbc6e6b729023531b7ee2d60b7a933fd4b65260469b948487112420","source":{"kind":"arxiv","id":"2605.17416","version":1},"attestation_state":"computed","paper":{"title":"Benchmarking Mythos-Linked Bug Rediscovery","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Even when supplied with the exact target source files, large language models rediscover only six of the intended Mythos-linked systems bugs across 54 attempts.","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Arthur Gervais, Isaac David","submitted_at":"2026-05-17T12:24:11Z","abstract_excerpt":"Anthropic's April 2026 Mythos materials combine benchmark claims with concrete bug-finding stories across OpenBSD, FreeBSD, Linux, FFmpeg, and browsers. This paper reports a controlled target-file rediscovery experiment on six public or high-confidence Mythos-linked systems tasks. Each model receives the same target file or files, read-only source tools, three repeats per task, and one manual target-matching rubric; prompts omit CVE identifiers, patch hashes, advisory text, author names, disclosure dates, and answer key root cause language. The experiment contains 54 counted model-task attempt"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.17416","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-05-17T12:24:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5b4970652583941957cfafa8246885efc42f035cc19195e08dc4407050e0c755","abstract_canon_sha256":"b34a9ab559bfbe0809b5076fad39b88a247e809cca6c95fde4dbc9bb224f482b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:57.343387Z","signature_b64":"PeDCscWjNgTi80oijjSg6W54ljTU7scPqOL9N5RisRS1yOjwEmS9Bg0vFYGNFi7MmfOL2fbAoIS0fo6GKF35Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be346dabbcbc6e6b729023531b7ee2d60b7a933fd4b65260469b948487112420","last_reissued_at":"2026-05-20T00:03:57.341680Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:57.341680Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Benchmarking Mythos-Linked Bug Rediscovery","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Even when supplied with the exact target source files, large language models rediscover only six of the intended Mythos-linked systems bugs across 54 attempts.","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Arthur Gervais, Isaac David","submitted_at":"2026-05-17T12:24:11Z","abstract_excerpt":"Anthropic's April 2026 Mythos materials combine benchmark claims with concrete bug-finding stories across OpenBSD, FreeBSD, Linux, FFmpeg, and browsers. This paper reports a controlled target-file rediscovery experiment on six public or high-confidence Mythos-linked systems tasks. Each model receives the same target file or files, read-only source tools, three repeats per task, and one manual target-matching rubric; prompts omit CVE identifiers, patch hashes, advisory text, author names, disclosure dates, and answer key root cause language. The experiment contains 54 counted model-task attempt"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Under this favorable target-file scaffold, systems-specific prompting yields only six target matches across 54 counted attempts.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The manual target-matching rubric correctly identifies rediscovery of the intended Mythos bug without bias from the specific language or details omitted from the prompts.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A benchmarking experiment finds low rediscovery rates for three models on six Mythos-linked bug tasks, with only six target matches across 54 attempts under controlled prompting.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Even when supplied with the exact target source files, large language models rediscover only six of the intended Mythos-linked systems bugs across 54 attempts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8bc23607f6ae3c44af980564a5d0efed0dd77c0bf6791d402556fbd233608c99"},"source":{"id":"2605.17416","kind":"arxiv","version":1},"verdict":{"id":"a08a541d-12ed-4c47-82f6-d3d932f7fa11","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T23:11:04.576523Z","strongest_claim":"Under this favorable target-file scaffold, systems-specific prompting yields only six target matches across 54 counted attempts.","one_line_summary":"A benchmarking experiment finds low rediscovery rates for three models on six Mythos-linked bug tasks, with only six target matches across 54 attempts under controlled prompting.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The manual target-matching rubric correctly identifies rediscovery of the intended Mythos bug without bias from the specific language or details omitted from the prompts.","pith_extraction_headline":"Even when supplied with the exact target source files, large language models rediscover only six of the intended Mythos-linked systems bugs across 54 attempts."},"integrity":{"clean":false,"summary":{"advisory":1,"critical":0,"by_detector":{"doi_compliance":{"total":1,"advisory":1,"critical":0,"informational":0}},"informational":0},"endpoint":"/pith/2605.17416/integrity.json","findings":[{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TDSC.2021.3051525.11) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":30,"audited_at":"2026-05-19T23:21:01.317386Z","detected_doi":"10.1109/TDSC.2021.3051525.11","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T23:31:19.978554Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T23:21:01.317386Z","status":"completed","version":"1.0.0","findings_count":1},{"name":"claim_evidence","ran_at":"2026-05-19T21:41:57.741798Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.687165Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"61fba97be5b9b0e639fa338d9da22bc687200000f9dcfa68c8475ae2863bf183"},"references":{"count":55,"sample":[{"doi":"","year":2026,"title":"Partnering with mozilla to improve firefox’s security, 2026","work_id":"701e45ac-d338-45f4-a344-c82060f08323","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Anthropic. Project glasswing, 2026. URL https://www.anthropic.com/glasswing. Ac- cessed 2026-04-18","work_id":"1f8988a9-4dcd-4c22-a8ca-95feea19fc83","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Claude mythos preview system card, 2026","work_id":"369aa627-6d39-42c5-b6ce-2db6bcccffb5","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Assessing claude mythos preview’s cybersecurity capabilities,","work_id":"c886aa75-9505-4523-b5a5-3312a0281eef","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Published 2026-04- 07; accessed 2026-04-18","work_id":"ef53a326-086f-4392-ba25-64a01a748b7d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":55,"snapshot_sha256":"35f8ea7a036d81b0df234a7200d182f9fe12707271125e20f81fc5c7df45b720","internal_anchors":2},"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":"2605.17416","created_at":"2026-05-20T00:03:57.341833+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17416v1","created_at":"2026-05-20T00:03:57.341833+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17416","created_at":"2026-05-20T00:03:57.341833+00:00"},{"alias_kind":"pith_short_12","alias_value":"XY2G3K54XRXG","created_at":"2026-05-20T00:03:57.341833+00:00"},{"alias_kind":"pith_short_16","alias_value":"XY2G3K54XRXGW4UQ","created_at":"2026-05-20T00:03:57.341833+00:00"},{"alias_kind":"pith_short_8","alias_value":"XY2G3K54","created_at":"2026-05-20T00:03:57.341833+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y","json":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y.json","graph_json":"https://pith.science/api/pith-number/XY2G3K54XRXGW4UQENJRW7XC2Y/graph.json","events_json":"https://pith.science/api/pith-number/XY2G3K54XRXGW4UQENJRW7XC2Y/events.json","paper":"https://pith.science/paper/XY2G3K54"},"agent_actions":{"view_html":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y","download_json":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y.json","view_paper":"https://pith.science/paper/XY2G3K54","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17416&json=true","fetch_graph":"https://pith.science/api/pith-number/XY2G3K54XRXGW4UQENJRW7XC2Y/graph.json","fetch_events":"https://pith.science/api/pith-number/XY2G3K54XRXGW4UQENJRW7XC2Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y/action/storage_attestation","attest_author":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y/action/author_attestation","sign_citation":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y/action/citation_signature","submit_replication":"https://pith.science/pith/XY2G3K54XRXGW4UQENJRW7XC2Y/action/replication_record"}},"created_at":"2026-05-20T00:03:57.341833+00:00","updated_at":"2026-05-20T00:03:57.341833+00:00"}