{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VBZ37CDRQCY33AH57HY3QOCQHS","short_pith_number":"pith:VBZ37CDR","schema_version":"1.0","canonical_sha256":"a873bf887180b1bd80fdf9f1b838503c8b8437645d7fceb0d55f6a5168e7a552","source":{"kind":"arxiv","id":"2510.23497","version":3},"attestation_state":"computed","paper":{"title":"VOLD: Reasoning Transfer from LLMs to Vision-Language Models via On-Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cordelia Schmid, Hilde Kuehne, Walid Bousselham","submitted_at":"2025-10-27T16:32:12Z","abstract_excerpt":"Training vision-language models (VLMs) for complex reasoning remains a challenging task, i.a. due to the scarcity of high-quality image-text reasoning data. Conversely, text-based reasoning resources are abundant and scalable, but it is still an open question how to leveraging them for VLM reasoning. To address this problem, we propose VOLD, a framework to transfer reasoning capabilities from text-only teacher models to VLM student models. To this end, VOLD combines reinforcement learning via Group Relative Policy Optimization (GRPO) with on-policy distillation, which allows the student reason"},"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":"2510.23497","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-10-27T16:32:12Z","cross_cats_sorted":[],"title_canon_sha256":"cfb6c4e6a44fef97354c3965ebecaff161bd2ae498640e9fe056bbdde2fc7304","abstract_canon_sha256":"b73ec25533503fb3756ee84791f254df2f0cfb28240a598c99140791b86c2112"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T00:13:44.140079Z","signature_b64":"2R0po+DVl4sKUXBBuqhyh6MsHGT1ZxjZkrMWNhUq0rfJhnI7RDkIkAZXRetKTR5gYbuLuklR9NtJx+xRvQKICw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a873bf887180b1bd80fdf9f1b838503c8b8437645d7fceb0d55f6a5168e7a552","last_reissued_at":"2026-06-05T00:13:44.139186Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T00:13:44.139186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VOLD: Reasoning Transfer from LLMs to Vision-Language Models via On-Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cordelia Schmid, Hilde Kuehne, Walid Bousselham","submitted_at":"2025-10-27T16:32:12Z","abstract_excerpt":"Training vision-language models (VLMs) for complex reasoning remains a challenging task, i.a. due to the scarcity of high-quality image-text reasoning data. Conversely, text-based reasoning resources are abundant and scalable, but it is still an open question how to leveraging them for VLM reasoning. To address this problem, we propose VOLD, a framework to transfer reasoning capabilities from text-only teacher models to VLM student models. To this end, VOLD combines reinforcement learning via Group Relative Policy Optimization (GRPO) with on-policy distillation, which allows the student reason"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.23497","kind":"arxiv","version":3},"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/2510.23497/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":"2510.23497","created_at":"2026-06-05T00:13:44.139304+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.23497v3","created_at":"2026-06-05T00:13:44.139304+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.23497","created_at":"2026-06-05T00:13:44.139304+00:00"},{"alias_kind":"pith_short_12","alias_value":"VBZ37CDRQCY3","created_at":"2026-06-05T00:13:44.139304+00:00"},{"alias_kind":"pith_short_16","alias_value":"VBZ37CDRQCY33AH5","created_at":"2026-06-05T00:13:44.139304+00:00"},{"alias_kind":"pith_short_8","alias_value":"VBZ37CDR","created_at":"2026-06-05T00:13:44.139304+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.13643","citing_title":"Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03677","citing_title":"Uni-OPD: Unifying On-Policy Distillation with a Dual-Perspective Recipe","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07725","citing_title":"SOD: Step-wise On-policy Distillation for Small Language Model Agents","ref_index":33,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS","json":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS.json","graph_json":"https://pith.science/api/pith-number/VBZ37CDRQCY33AH57HY3QOCQHS/graph.json","events_json":"https://pith.science/api/pith-number/VBZ37CDRQCY33AH57HY3QOCQHS/events.json","paper":"https://pith.science/paper/VBZ37CDR"},"agent_actions":{"view_html":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS","download_json":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS.json","view_paper":"https://pith.science/paper/VBZ37CDR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.23497&json=true","fetch_graph":"https://pith.science/api/pith-number/VBZ37CDRQCY33AH57HY3QOCQHS/graph.json","fetch_events":"https://pith.science/api/pith-number/VBZ37CDRQCY33AH57HY3QOCQHS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS/action/storage_attestation","attest_author":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS/action/author_attestation","sign_citation":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS/action/citation_signature","submit_replication":"https://pith.science/pith/VBZ37CDRQCY33AH57HY3QOCQHS/action/replication_record"}},"created_at":"2026-06-05T00:13:44.139304+00:00","updated_at":"2026-06-05T00:13:44.139304+00:00"}