{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WWLJPLPXEEBNMPC6BHI2ATDOCC","short_pith_number":"pith:WWLJPLPX","schema_version":"1.0","canonical_sha256":"b59697adf72102d63c5e09d1a04c6e10b40e3fc4407aa3d855518f1948506a40","source":{"kind":"arxiv","id":"2606.19120","version":1},"attestation_state":"computed","paper":{"title":"Seeing Before Reasoning: Decoupling Perception and Reasoning for Shortcut-Resilient Multimodal On-Policy Self-Distillation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Lianqing Liu, Sihan Wang, Xiyao Liu, Zhi Han","submitted_at":"2026-06-17T14:33:38Z","abstract_excerpt":"On-policy self-distillation (OPSD) trains a model on its own rollouts and uses a frozen copy to provide dense token-level targets conditioned on a reference target. This works well for LLM reasoning, but a direct extension to multimodal large language models (MLLMs) can create a shortcut: the privileged target may guide tokens mainly based on the text reference target rather than the image. We propose ViGOS, a visually grounded OPSD framework for MLLM post-training. The student first writes a visual description and then reasons toward the final answer. For valid rollouts, an image-only percept"},"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":"2606.19120","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-17T14:33:38Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"c958243313567fc1b05504b2413559cc1b9c3ce61d2b500b1c572f8a3d4c4e5e","abstract_canon_sha256":"0b48115373e88ae8903034d80a8792dad01bc3ce848ac2e58b999ab397cb5d29"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:11:57.165521Z","signature_b64":"m0GM3N7eq8jRv4U+kiKVBqbWgoyaWp2FQpB2ev4HpZ7UfgAGJYLzWmOyQbiW6vc0Ah6V7+zhLQMp6F1shYgsAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b59697adf72102d63c5e09d1a04c6e10b40e3fc4407aa3d855518f1948506a40","last_reissued_at":"2026-06-19T16:11:57.165157Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:11:57.165157Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Seeing Before Reasoning: Decoupling Perception and Reasoning for Shortcut-Resilient Multimodal On-Policy Self-Distillation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Lianqing Liu, Sihan Wang, Xiyao Liu, Zhi Han","submitted_at":"2026-06-17T14:33:38Z","abstract_excerpt":"On-policy self-distillation (OPSD) trains a model on its own rollouts and uses a frozen copy to provide dense token-level targets conditioned on a reference target. This works well for LLM reasoning, but a direct extension to multimodal large language models (MLLMs) can create a shortcut: the privileged target may guide tokens mainly based on the text reference target rather than the image. We propose ViGOS, a visually grounded OPSD framework for MLLM post-training. The student first writes a visual description and then reasons toward the final answer. For valid rollouts, an image-only percept"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.19120","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/2606.19120/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":"2606.19120","created_at":"2026-06-19T16:11:57.165218+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.19120v1","created_at":"2026-06-19T16:11:57.165218+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.19120","created_at":"2026-06-19T16:11:57.165218+00:00"},{"alias_kind":"pith_short_12","alias_value":"WWLJPLPXEEBN","created_at":"2026-06-19T16:11:57.165218+00:00"},{"alias_kind":"pith_short_16","alias_value":"WWLJPLPXEEBNMPC6","created_at":"2026-06-19T16:11:57.165218+00:00"},{"alias_kind":"pith_short_8","alias_value":"WWLJPLPX","created_at":"2026-06-19T16:11:57.165218+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/WWLJPLPXEEBNMPC6BHI2ATDOCC","json":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC.json","graph_json":"https://pith.science/api/pith-number/WWLJPLPXEEBNMPC6BHI2ATDOCC/graph.json","events_json":"https://pith.science/api/pith-number/WWLJPLPXEEBNMPC6BHI2ATDOCC/events.json","paper":"https://pith.science/paper/WWLJPLPX"},"agent_actions":{"view_html":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC","download_json":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC.json","view_paper":"https://pith.science/paper/WWLJPLPX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.19120&json=true","fetch_graph":"https://pith.science/api/pith-number/WWLJPLPXEEBNMPC6BHI2ATDOCC/graph.json","fetch_events":"https://pith.science/api/pith-number/WWLJPLPXEEBNMPC6BHI2ATDOCC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC/action/storage_attestation","attest_author":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC/action/author_attestation","sign_citation":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC/action/citation_signature","submit_replication":"https://pith.science/pith/WWLJPLPXEEBNMPC6BHI2ATDOCC/action/replication_record"}},"created_at":"2026-06-19T16:11:57.165218+00:00","updated_at":"2026-06-19T16:11:57.165218+00:00"}