{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VC5XR6NDM2XCJ3SSWLNZ45HXOV","short_pith_number":"pith:VC5XR6ND","schema_version":"1.0","canonical_sha256":"a8bb78f9a366ae24ee52b2db9e74f7756cfc4b29f4499349444a679f2fbb4be6","source":{"kind":"arxiv","id":"2605.20177","version":1},"attestation_state":"computed","paper":{"title":"From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Cihang Xie, Freda Shi, Hanqing Lu, Haoqin Tu, Hardy Chen, Hui Liu, Juncheng Wu, Xianfeng Tang, Yuyin Zhou","submitted_at":"2026-05-19T17:58:40Z","abstract_excerpt":"Recent advances in vision-language models (VLMs) emphasize long chain-of-thought reasoning; yet, we find that their performance on visual tasks is primarily limited by a lack of visual perception as opposed to reasoning itself. In this work, we systematically study the interplay between perception and reasoning in VLM post-training by decomposing their capabilities into three separate training stages: visual perception, visual reasoning, and textual reasoning, incorporating specialized training data. We demonstrate that visual perception (a) requires targeted optimization with specialized data"},"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":"2605.20177","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-19T17:58:40Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"a10070eaf367e50fea04f65a7e478307f528e2a5fda74bf2261a79b9b78b51e0","abstract_canon_sha256":"2213cbd89b3e79e5ad300758e79f2d652126ca31e96b7bb3eb005603a267aede"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:06:06.424408Z","signature_b64":"GjFDA8fsA+u9AJ1OkMC85yf4onHSL4A1Hm/cUYPl36Av0G1phZqcrktC6TED5p6oFh5VcMtzSiG1IAU10B0zDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8bb78f9a366ae24ee52b2db9e74f7756cfc4b29f4499349444a679f2fbb4be6","last_reissued_at":"2026-05-20T02:06:06.423579Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:06:06.423579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Cihang Xie, Freda Shi, Hanqing Lu, Haoqin Tu, Hardy Chen, Hui Liu, Juncheng Wu, Xianfeng Tang, Yuyin Zhou","submitted_at":"2026-05-19T17:58:40Z","abstract_excerpt":"Recent advances in vision-language models (VLMs) emphasize long chain-of-thought reasoning; yet, we find that their performance on visual tasks is primarily limited by a lack of visual perception as opposed to reasoning itself. In this work, we systematically study the interplay between perception and reasoning in VLM post-training by decomposing their capabilities into three separate training stages: visual perception, visual reasoning, and textual reasoning, incorporating specialized training data. We demonstrate that visual perception (a) requires targeted optimization with specialized data"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20177","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/2605.20177/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":"2605.20177","created_at":"2026-05-20T02:06:06.423709+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.20177v1","created_at":"2026-05-20T02:06:06.423709+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20177","created_at":"2026-05-20T02:06:06.423709+00:00"},{"alias_kind":"pith_short_12","alias_value":"VC5XR6NDM2XC","created_at":"2026-05-20T02:06:06.423709+00:00"},{"alias_kind":"pith_short_16","alias_value":"VC5XR6NDM2XCJ3SS","created_at":"2026-05-20T02:06:06.423709+00:00"},{"alias_kind":"pith_short_8","alias_value":"VC5XR6ND","created_at":"2026-05-20T02:06:06.423709+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/VC5XR6NDM2XCJ3SSWLNZ45HXOV","json":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV.json","graph_json":"https://pith.science/api/pith-number/VC5XR6NDM2XCJ3SSWLNZ45HXOV/graph.json","events_json":"https://pith.science/api/pith-number/VC5XR6NDM2XCJ3SSWLNZ45HXOV/events.json","paper":"https://pith.science/paper/VC5XR6ND"},"agent_actions":{"view_html":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV","download_json":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV.json","view_paper":"https://pith.science/paper/VC5XR6ND","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.20177&json=true","fetch_graph":"https://pith.science/api/pith-number/VC5XR6NDM2XCJ3SSWLNZ45HXOV/graph.json","fetch_events":"https://pith.science/api/pith-number/VC5XR6NDM2XCJ3SSWLNZ45HXOV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV/action/storage_attestation","attest_author":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV/action/author_attestation","sign_citation":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV/action/citation_signature","submit_replication":"https://pith.science/pith/VC5XR6NDM2XCJ3SSWLNZ45HXOV/action/replication_record"}},"created_at":"2026-05-20T02:06:06.423709+00:00","updated_at":"2026-05-20T02:06:06.423709+00:00"}