{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:UFR2BXRU5E456SURIPLSQFXGUX","short_pith_number":"pith:UFR2BXRU","canonical_record":{"source":{"id":"2604.03447","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-03T20:38:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e4c95a618f4d421f830c2fab73754c0592c0781ce3253a69d79c3576b1dbce6f","abstract_canon_sha256":"f090aedc264eba67c88433b4d6fd447629943d15497275b362cd2625e6b9b015"},"schema_version":"1.0"},"canonical_sha256":"a163a0de34e939df4a9143d72816e6a5c42e7a08e8e4b0f16b926b74cc409b28","source":{"kind":"arxiv","id":"2604.03447","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.03447","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.03447v2","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.03447","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_12","alias_value":"UFR2BXRU5E45","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_16","alias_value":"UFR2BXRU5E456SUR","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_8","alias_value":"UFR2BXRU","created_at":"2026-07-23T00:23:39Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:UFR2BXRU5E456SURIPLSQFXGUX","target":"record","payload":{"canonical_record":{"source":{"id":"2604.03447","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-03T20:38:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e4c95a618f4d421f830c2fab73754c0592c0781ce3253a69d79c3576b1dbce6f","abstract_canon_sha256":"f090aedc264eba67c88433b4d6fd447629943d15497275b362cd2625e6b9b015"},"schema_version":"1.0"},"canonical_sha256":"a163a0de34e939df4a9143d72816e6a5c42e7a08e8e4b0f16b926b74cc409b28","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-23T00:23:39.657889Z","signature_b64":"8aeh+hjpoiP4GRzIGtYTseDh091AxXNjGPXqXUALfIzg7/8wxwmx/AGSWFhEbgf5gdGX5FGYGxuIuEGTwLv+BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a163a0de34e939df4a9143d72816e6a5c42e7a08e8e4b0f16b926b74cc409b28","last_reissued_at":"2026-07-23T00:23:39.656787Z","signature_status":"signed_v1","first_computed_at":"2026-07-23T00:23:39.656787Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.03447","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-23T00:23:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Jhr5XfRhfwH2x//4rZU7Hm8qLIv7OJkarQGwNFdvhfQFPoNfszm+Yq04fh8OGzFjknsCjlNOfj7aXIxd0c+JBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-01T21:08:48.144422Z"},"content_sha256":"8509eb2daef95af57905b8e4d3ab28fdbb7d62b90c0d5b68cfb551ad3767b6ab","schema_version":"1.0","event_id":"sha256:8509eb2daef95af57905b8e4d3ab28fdbb7d62b90c0d5b68cfb551ad3767b6ab"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:UFR2BXRU5E456SURIPLSQFXGUX","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Measuring LLM Trust Allocation Across Conflicting Software Artifacts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible.","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Ahsanul Ameen Sabit, Noshin Ulfat, Soneya Binta Hossain","submitted_at":"2026-04-03T20:38:14Z","abstract_excerpt":"LLM-based software engineering assistants often reason over multiple artifacts, including code, documentation, signatures, and tests, even when those artifacts are incomplete or mutually inconsistent. Existing evaluations primarily measure final outputs, leaving unclear whether a model recognized unreliable evidence, identified the faulty source, or prioritized the appropriate artifact.\n  We introduce TRACE, a controlled method for evaluating how LLMs assess and prioritize conflicting software artifacts. TRACE constructs paired clean and perturbed versions of real-world Java method bundles by "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across all models, quality penalties are largely localized to the perturbed artifact and increase with severity, but sensitivity is asymmetric across artifact types: documentation bugs induce a substantially larger heavy-to-subtle gap than implementation faults (0.152-0.253 vs. 0.049-0.123). Models detect explicit documentation bugs well (67-94%) and Javadoc and implementation contradictions at 50-91%, yet show a systematic blind spot when only the implementation drifts while the documentation remains plausible, with detection dropping by 7-42 percentage points.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the curated Java method bundles and blind perturbations sufficiently represent real-world conflicting artifacts and that the elicited trust traces accurately capture internal model reasoning rather than surface-level pattern matching.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"TRACE reveals that LLMs detect documentation bugs and contradictions better than subtle implementation drift, with asymmetric sensitivity and poor confidence calibration across seven models on 22k traces.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"718f8dd91d7611f7affbc094b922823348c6b58df6c6c4d3912ad62e73c3d4e6"},"source":{"id":"2604.03447","kind":"arxiv","version":2},"verdict":{"id":"cbb13541-349c-4ef1-9038-2d11014998cc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T18:10:53.482492Z","strongest_claim":"Across all models, quality penalties are largely localized to the perturbed artifact and increase with severity, but sensitivity is asymmetric across artifact types: documentation bugs induce a substantially larger heavy-to-subtle gap than implementation faults (0.152-0.253 vs. 0.049-0.123). Models detect explicit documentation bugs well (67-94%) and Javadoc and implementation contradictions at 50-91%, yet show a systematic blind spot when only the implementation drifts while the documentation remains plausible, with detection dropping by 7-42 percentage points.","one_line_summary":"TRACE reveals that LLMs detect documentation bugs and contradictions better than subtle implementation drift, with asymmetric sensitivity and poor confidence calibration across seven models on 22k traces.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the curated Java method bundles and blind perturbations sufficiently represent real-world conflicting artifacts and that the elicited trust traces accurately capture internal model reasoning rather than surface-level pattern matching.","pith_extraction_headline":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.03447/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":2,"snapshot_sha256":"b98d827d4d8b93ded7dee13f62a1d482c34b14776ecd9812355e4d2bf8f64345"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"cbb13541-349c-4ef1-9038-2d11014998cc"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-23T00:23:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qtijzHS3808nzmRKlOGgGGxS+UA3w0ns3AfsxhHmJqdGkOiU/rX0RSdkTsETn3M7pbBLLF46OinqsI4FuxOFAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-01T21:08:48.145183Z"},"content_sha256":"c57d866e183c449dc384649dc5f2109bf600d89a918a1fd4130e1b65bed756aa","schema_version":"1.0","event_id":"sha256:c57d866e183c449dc384649dc5f2109bf600d89a918a1fd4130e1b65bed756aa"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UFR2BXRU5E456SURIPLSQFXGUX/bundle.json","state_url":"https://pith.science/pith/UFR2BXRU5E456SURIPLSQFXGUX/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UFR2BXRU5E456SURIPLSQFXGUX/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-01T21:08:48Z","links":{"resolver":"https://pith.science/pith/UFR2BXRU5E456SURIPLSQFXGUX","bundle":"https://pith.science/pith/UFR2BXRU5E456SURIPLSQFXGUX/bundle.json","state":"https://pith.science/pith/UFR2BXRU5E456SURIPLSQFXGUX/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UFR2BXRU5E456SURIPLSQFXGUX/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:UFR2BXRU5E456SURIPLSQFXGUX","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f090aedc264eba67c88433b4d6fd447629943d15497275b362cd2625e6b9b015","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-03T20:38:14Z","title_canon_sha256":"e4c95a618f4d421f830c2fab73754c0592c0781ce3253a69d79c3576b1dbce6f"},"schema_version":"1.0","source":{"id":"2604.03447","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.03447","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.03447v2","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.03447","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_12","alias_value":"UFR2BXRU5E45","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_16","alias_value":"UFR2BXRU5E456SUR","created_at":"2026-07-23T00:23:39Z"},{"alias_kind":"pith_short_8","alias_value":"UFR2BXRU","created_at":"2026-07-23T00:23:39Z"}],"graph_snapshots":[{"event_id":"sha256:c57d866e183c449dc384649dc5f2109bf600d89a918a1fd4130e1b65bed756aa","target":"graph","created_at":"2026-07-23T00:23:39Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Across all models, quality penalties are largely localized to the perturbed artifact and increase with severity, but sensitivity is asymmetric across artifact types: documentation bugs induce a substantially larger heavy-to-subtle gap than implementation faults (0.152-0.253 vs. 0.049-0.123). Models detect explicit documentation bugs well (67-94%) and Javadoc and implementation contradictions at 50-91%, yet show a systematic blind spot when only the implementation drifts while the documentation remains plausible, with detection dropping by 7-42 percentage points."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the curated Java method bundles and blind perturbations sufficiently represent real-world conflicting artifacts and that the elicited trust traces accurately capture internal model reasoning rather than surface-level pattern matching."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"TRACE reveals that LLMs detect documentation bugs and contradictions better than subtle implementation drift, with asymmetric sensitivity and poor confidence calibration across seven models on 22k traces."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible."}],"snapshot_sha256":"718f8dd91d7611f7affbc094b922823348c6b58df6c6c4d3912ad62e73c3d4e6"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"b98d827d4d8b93ded7dee13f62a1d482c34b14776ecd9812355e4d2bf8f64345"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.03447/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"LLM-based software engineering assistants often reason over multiple artifacts, including code, documentation, signatures, and tests, even when those artifacts are incomplete or mutually inconsistent. Existing evaluations primarily measure final outputs, leaving unclear whether a model recognized unreliable evidence, identified the faulty source, or prioritized the appropriate artifact.\n  We introduce TRACE, a controlled method for evaluating how LLMs assess and prioritize conflicting software artifacts. TRACE constructs paired clean and perturbed versions of real-world Java method bundles by ","authors_text":"Ahsanul Ameen Sabit, Noshin Ulfat, Soneya Binta Hossain","cross_cats":["cs.AI"],"headline":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-03T20:38:14Z","title":"Measuring LLM Trust Allocation Across Conflicting Software Artifacts"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.03447","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-13T18:10:53.482492Z","id":"cbb13541-349c-4ef1-9038-2d11014998cc","model_set":{"reader":"grok-4.3"},"one_line_summary":"TRACE reveals that LLMs detect documentation bugs and contradictions better than subtle implementation drift, with asymmetric sensitivity and poor confidence calibration across seven models on 22k traces.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"LLMs penalize documentation bugs more than implementation faults when artifacts conflict, but overlook code drift when docs remain plausible.","strongest_claim":"Across all models, quality penalties are largely localized to the perturbed artifact and increase with severity, but sensitivity is asymmetric across artifact types: documentation bugs induce a substantially larger heavy-to-subtle gap than implementation faults (0.152-0.253 vs. 0.049-0.123). Models detect explicit documentation bugs well (67-94%) and Javadoc and implementation contradictions at 50-91%, yet show a systematic blind spot when only the implementation drifts while the documentation remains plausible, with detection dropping by 7-42 percentage points.","weakest_assumption":"That the curated Java method bundles and blind perturbations sufficiently represent real-world conflicting artifacts and that the elicited trust traces accurately capture internal model reasoning rather than surface-level pattern matching."}},"verdict_id":"cbb13541-349c-4ef1-9038-2d11014998cc"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:8509eb2daef95af57905b8e4d3ab28fdbb7d62b90c0d5b68cfb551ad3767b6ab","target":"record","created_at":"2026-07-23T00:23:39Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f090aedc264eba67c88433b4d6fd447629943d15497275b362cd2625e6b9b015","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-04-03T20:38:14Z","title_canon_sha256":"e4c95a618f4d421f830c2fab73754c0592c0781ce3253a69d79c3576b1dbce6f"},"schema_version":"1.0","source":{"id":"2604.03447","kind":"arxiv","version":2}},"canonical_sha256":"a163a0de34e939df4a9143d72816e6a5c42e7a08e8e4b0f16b926b74cc409b28","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a163a0de34e939df4a9143d72816e6a5c42e7a08e8e4b0f16b926b74cc409b28","first_computed_at":"2026-07-23T00:23:39.656787Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-23T00:23:39.656787Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"8aeh+hjpoiP4GRzIGtYTseDh091AxXNjGPXqXUALfIzg7/8wxwmx/AGSWFhEbgf5gdGX5FGYGxuIuEGTwLv+BA==","signature_status":"signed_v1","signed_at":"2026-07-23T00:23:39.657889Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.03447","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8509eb2daef95af57905b8e4d3ab28fdbb7d62b90c0d5b68cfb551ad3767b6ab","sha256:c57d866e183c449dc384649dc5f2109bf600d89a918a1fd4130e1b65bed756aa"],"state_sha256":"3470b2ea996d764f81849169273c728c2b6de11d31e1a8787f2bde86d3a80f16"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XNZSJd1FrhOfydqs9LSYc3bEt6qz1+XVC6bDFy2I/WJAAtVSp9RWHflqs0lhZb6kCJiNcA9MzaRdyS7qUfCwDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-01T21:08:48.151895Z","bundle_sha256":"8dda8878261d75cd03ea0f27e389204d4ce011f4657fe355f09ff0f36f8014e1"}}