{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3LOHOJXSW5QCLE6N5ZE4DOJLU3","short_pith_number":"pith:3LOHOJXS","schema_version":"1.0","canonical_sha256":"dadc7726f2b7602593cdee49c1b92ba6eba3160d4aefdda68ce21769c2d3f0d7","source":{"kind":"arxiv","id":"2605.29496","version":1},"attestation_state":"computed","paper":{"title":"On Asymmetric Optimization of Reasoning and Perception in Vision-Language Model Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Kai-Wei Chang, Nanyun Peng, Xueqing Wu, Yu-Chi Lin","submitted_at":"2026-05-28T07:20:49Z","abstract_excerpt":"Post-training has greatly improved reasoning in frontier vision-language models, yet its gains for perception remain comparatively limited, creating a bottleneck for end-to-end visual reasoning. To investigate this gap, we introduce a controlled diagnostic framework with two synthetic tasks that disentangle perception from reasoning. Our analysis reveals a consistent perception-reasoning asymmetry: posttraining improves reasoning more substantially than perception, though the underlying mechanism differs by training paradigm. For supervised fine-tuning (SFT), this asymmetry stems from token im"},"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":"2605.29496","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-28T07:20:49Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"d02f2912182669ffbdd4666ea53bd128502fe43987fe883493912e9f4b3a35ae","abstract_canon_sha256":"8cd6cc29c0c6345b09f8c6d99359dc3dbf7ba428a6ee53a06fdbde26d1e6f368"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:42.789960Z","signature_b64":"+k2km2OjUkYQp/Kyp3Tric7Z2T0Xyqy/i1b6m6OECPs4n3wHsz3QPpY7htqx9qvqDgHZcLl3n2MISBMoPMZeBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dadc7726f2b7602593cdee49c1b92ba6eba3160d4aefdda68ce21769c2d3f0d7","last_reissued_at":"2026-05-29T01:05:42.789063Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:42.789063Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Asymmetric Optimization of Reasoning and Perception in Vision-Language Model Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Kai-Wei Chang, Nanyun Peng, Xueqing Wu, Yu-Chi Lin","submitted_at":"2026-05-28T07:20:49Z","abstract_excerpt":"Post-training has greatly improved reasoning in frontier vision-language models, yet its gains for perception remain comparatively limited, creating a bottleneck for end-to-end visual reasoning. To investigate this gap, we introduce a controlled diagnostic framework with two synthetic tasks that disentangle perception from reasoning. Our analysis reveals a consistent perception-reasoning asymmetry: posttraining improves reasoning more substantially than perception, though the underlying mechanism differs by training paradigm. For supervised fine-tuning (SFT), this asymmetry stems from token im"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29496","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/2605.29496/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":"2605.29496","created_at":"2026-05-29T01:05:42.789172+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29496v1","created_at":"2026-05-29T01:05:42.789172+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29496","created_at":"2026-05-29T01:05:42.789172+00:00"},{"alias_kind":"pith_short_12","alias_value":"3LOHOJXSW5QC","created_at":"2026-05-29T01:05:42.789172+00:00"},{"alias_kind":"pith_short_16","alias_value":"3LOHOJXSW5QCLE6N","created_at":"2026-05-29T01:05:42.789172+00:00"},{"alias_kind":"pith_short_8","alias_value":"3LOHOJXS","created_at":"2026-05-29T01:05:42.789172+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/3LOHOJXSW5QCLE6N5ZE4DOJLU3","json":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3.json","graph_json":"https://pith.science/api/pith-number/3LOHOJXSW5QCLE6N5ZE4DOJLU3/graph.json","events_json":"https://pith.science/api/pith-number/3LOHOJXSW5QCLE6N5ZE4DOJLU3/events.json","paper":"https://pith.science/paper/3LOHOJXS"},"agent_actions":{"view_html":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3","download_json":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3.json","view_paper":"https://pith.science/paper/3LOHOJXS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29496&json=true","fetch_graph":"https://pith.science/api/pith-number/3LOHOJXSW5QCLE6N5ZE4DOJLU3/graph.json","fetch_events":"https://pith.science/api/pith-number/3LOHOJXSW5QCLE6N5ZE4DOJLU3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3/action/storage_attestation","attest_author":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3/action/author_attestation","sign_citation":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3/action/citation_signature","submit_replication":"https://pith.science/pith/3LOHOJXSW5QCLE6N5ZE4DOJLU3/action/replication_record"}},"created_at":"2026-05-29T01:05:42.789172+00:00","updated_at":"2026-05-29T01:05:42.789172+00:00"}