{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:36COOF52A5Y3C4IZPN2POEOZLW","short_pith_number":"pith:36COOF52","schema_version":"1.0","canonical_sha256":"df84e717ba0771b171197b74f711d95d870160024c021ec82b34294b515aac60","source":{"kind":"arxiv","id":"2606.01558","version":1},"attestation_state":"computed","paper":{"title":"Attention-guided Fine-tuning of Multimodal Large Language Models Improves Chain-of-Thought Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Aidong Zhang, Bohan Liu, Guangzhi Xiong, Sanchit Sinha, Zhenghao He","submitted_at":"2026-06-01T02:02:23Z","abstract_excerpt":"The effectiveness of Chain-of-Thought (CoT) prompting in Multimodal Large Language Models (MLLMs) remains uncertain: across several visual reasoning benchmarks, CoT prompting often degrades performance compared to direct prompting. In this paper, we provide a systematic analysis of CoT behavior in three modern MLLM families across model scales on datasets requiring step-wise visual evidence. Our analysis identifies two recurring failure modes: premature answer commitment and limited direct visual-token access during rationale generation. We further find that standard CoT-style Supervised Fine-"},"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":"2606.01558","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-06-01T02:02:23Z","cross_cats_sorted":[],"title_canon_sha256":"9fa27e6474ddfaf381990dd96e46a5763c13e57b11c03f15485aec287c24941e","abstract_canon_sha256":"91a234ba7ad34b1f3d737158b9307cce512bf67e8db8669b7b7f34e2d36fc52b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T02:04:36.458866Z","signature_b64":"qoekP7EcKJHh/yBpUSFl0IBUFE/Ht4OTCjbHA5KjLznjQumXXqo/cvpK/58VMsKrret2vPA+RQ9Sj1yJnoI5DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"df84e717ba0771b171197b74f711d95d870160024c021ec82b34294b515aac60","last_reissued_at":"2026-06-02T02:04:36.458451Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T02:04:36.458451Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention-guided Fine-tuning of Multimodal Large Language Models Improves Chain-of-Thought Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Aidong Zhang, Bohan Liu, Guangzhi Xiong, Sanchit Sinha, Zhenghao He","submitted_at":"2026-06-01T02:02:23Z","abstract_excerpt":"The effectiveness of Chain-of-Thought (CoT) prompting in Multimodal Large Language Models (MLLMs) remains uncertain: across several visual reasoning benchmarks, CoT prompting often degrades performance compared to direct prompting. In this paper, we provide a systematic analysis of CoT behavior in three modern MLLM families across model scales on datasets requiring step-wise visual evidence. Our analysis identifies two recurring failure modes: premature answer commitment and limited direct visual-token access during rationale generation. We further find that standard CoT-style Supervised Fine-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.01558","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/2606.01558/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":"2606.01558","created_at":"2026-06-02T02:04:36.458513+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.01558v1","created_at":"2026-06-02T02:04:36.458513+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.01558","created_at":"2026-06-02T02:04:36.458513+00:00"},{"alias_kind":"pith_short_12","alias_value":"36COOF52A5Y3","created_at":"2026-06-02T02:04:36.458513+00:00"},{"alias_kind":"pith_short_16","alias_value":"36COOF52A5Y3C4IZ","created_at":"2026-06-02T02:04:36.458513+00:00"},{"alias_kind":"pith_short_8","alias_value":"36COOF52","created_at":"2026-06-02T02:04:36.458513+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/36COOF52A5Y3C4IZPN2POEOZLW","json":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW.json","graph_json":"https://pith.science/api/pith-number/36COOF52A5Y3C4IZPN2POEOZLW/graph.json","events_json":"https://pith.science/api/pith-number/36COOF52A5Y3C4IZPN2POEOZLW/events.json","paper":"https://pith.science/paper/36COOF52"},"agent_actions":{"view_html":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW","download_json":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW.json","view_paper":"https://pith.science/paper/36COOF52","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.01558&json=true","fetch_graph":"https://pith.science/api/pith-number/36COOF52A5Y3C4IZPN2POEOZLW/graph.json","fetch_events":"https://pith.science/api/pith-number/36COOF52A5Y3C4IZPN2POEOZLW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW/action/storage_attestation","attest_author":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW/action/author_attestation","sign_citation":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW/action/citation_signature","submit_replication":"https://pith.science/pith/36COOF52A5Y3C4IZPN2POEOZLW/action/replication_record"}},"created_at":"2026-06-02T02:04:36.458513+00:00","updated_at":"2026-06-02T02:04:36.458513+00:00"}