{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BS276OFM4IOJ22QOZE23TAKC47","short_pith_number":"pith:BS276OFM","schema_version":"1.0","canonical_sha256":"0cb5ff38ace21c9d6a0ec935b98142e7f3ec581422aa57f5c0f87f8adaeecdfb","source":{"kind":"arxiv","id":"2608.11727","version":1},"attestation_state":"computed","paper":{"title":"Harness-IF: Evaluating Instruction Following Across Instruction Surfaces in Coding Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Changxin Pu, Ge Zhang, Haodong Wang, Haoran Que, Hong Zeng, Jin Chen, Shen Yan, Wenhao Huang, Zhongfei Hou, Zining Huang, Zuo Wang","submitted_at":"2026-08-12T07:07:57Z","abstract_excerpt":"When a coding agent obeys a rule, it may simply have been going to do that anyway. Existing instruction-following benchmarks cannot tell the difference: they concentrate rules in the user turn, while coding-agent benchmarks emphasize final task success. We introduce Harness-IF, which scores operational rules one at a time from execution evidence: 60 realistic multi-turn coding items drawn from a 642-rule library, 256 rules receiving verdicts, placed on the five configurable surfaces a deployed agent reads. To separate compliance from coincidence we introduce Against-Prior Accuracy (AP-Acc), wh"},"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":"2608.11727","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-08-12T07:07:57Z","cross_cats_sorted":[],"title_canon_sha256":"0b84f10a5e5042a199a8eb7b465bb95e0640583a86ff98e730e5f9e3f47bbc1b","abstract_canon_sha256":"3dcab8da46f072e2bcbe9d2b7f927f79bfb12c71f560c7ab7c96eee4f578e68d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-13T01:27:56.610118Z","signature_b64":"/k+MJdKVkwAbI5loLb2ikOzdhWdpmFoXt2r9fh5/3zkatERm3GKt2fVHbYZ22KIjce1SRQwv8jxmc9+GsTDoCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0cb5ff38ace21c9d6a0ec935b98142e7f3ec581422aa57f5c0f87f8adaeecdfb","last_reissued_at":"2026-08-13T01:27:56.607791Z","signature_status":"signed_v1","first_computed_at":"2026-08-13T01:27:56.607791Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Harness-IF: Evaluating Instruction Following Across Instruction Surfaces in Coding Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Changxin Pu, Ge Zhang, Haodong Wang, Haoran Que, Hong Zeng, Jin Chen, Shen Yan, Wenhao Huang, Zhongfei Hou, Zining Huang, Zuo Wang","submitted_at":"2026-08-12T07:07:57Z","abstract_excerpt":"When a coding agent obeys a rule, it may simply have been going to do that anyway. Existing instruction-following benchmarks cannot tell the difference: they concentrate rules in the user turn, while coding-agent benchmarks emphasize final task success. We introduce Harness-IF, which scores operational rules one at a time from execution evidence: 60 realistic multi-turn coding items drawn from a 642-rule library, 256 rules receiving verdicts, placed on the five configurable surfaces a deployed agent reads. To separate compliance from coincidence we introduce Against-Prior Accuracy (AP-Acc), wh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.11727","kind":"arxiv","version":1},"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/2608.11727/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":"2608.11727","created_at":"2026-08-13T01:27:56.608881+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.11727v1","created_at":"2026-08-13T01:27:56.608881+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.11727","created_at":"2026-08-13T01:27:56.608881+00:00"},{"alias_kind":"pith_short_12","alias_value":"BS276OFM4IOJ","created_at":"2026-08-13T01:27:56.608881+00:00"},{"alias_kind":"pith_short_16","alias_value":"BS276OFM4IOJ22QO","created_at":"2026-08-13T01:27:56.608881+00:00"},{"alias_kind":"pith_short_8","alias_value":"BS276OFM","created_at":"2026-08-13T01:27:56.608881+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/BS276OFM4IOJ22QOZE23TAKC47","json":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47.json","graph_json":"https://pith.science/api/pith-number/BS276OFM4IOJ22QOZE23TAKC47/graph.json","events_json":"https://pith.science/api/pith-number/BS276OFM4IOJ22QOZE23TAKC47/events.json","paper":"https://pith.science/paper/BS276OFM"},"agent_actions":{"view_html":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47","download_json":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47.json","view_paper":"https://pith.science/paper/BS276OFM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.11727&json=true","fetch_graph":"https://pith.science/api/pith-number/BS276OFM4IOJ22QOZE23TAKC47/graph.json","fetch_events":"https://pith.science/api/pith-number/BS276OFM4IOJ22QOZE23TAKC47/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47/action/storage_attestation","attest_author":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47/action/author_attestation","sign_citation":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47/action/citation_signature","submit_replication":"https://pith.science/pith/BS276OFM4IOJ22QOZE23TAKC47/action/replication_record"}},"created_at":"2026-08-13T01:27:56.608881+00:00","updated_at":"2026-08-13T01:27:56.608881+00:00"}