{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:INYKEWFEURNSAAVE4AUCLTWDMU","short_pith_number":"pith:INYKEWFE","schema_version":"1.0","canonical_sha256":"4370a258a4a45b2002a4e02825cec36500c69acd3206c9239956417c60088bb0","source":{"kind":"arxiv","id":"2605.24794","version":1},"attestation_state":"computed","paper":{"title":"DUEL: Adversarial Self-Play for Multimodal Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bingjun Sun, Guangdeng Liao, Hanqing Zeng, Ji Liu, Lin Qiu, Yao Liu","submitted_at":"2026-05-24T00:38:26Z","abstract_excerpt":"Reinforcement learning (RL) has emerged as an effective paradigm for improving the reasoning capability of vision-language models (VLMs). However, RL-based optimization typically depends on costly high-quality annotations that are difficult to scale. Existing unsupervised alternatives may drift toward biased solutions due to weak visual grounding and the lack of reliable verification signals. We propose a self-evolving post-training framework, DUEL, where supervision emerges from adversarial interactions between two policies initialized from the same pretrained VLM. A Challenger generates an i"},"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":"2605.24794","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-24T00:38:26Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"e80bf54f68b596737541dc951b3055114f87d6026e7ce1ad1957050cae691c1e","abstract_canon_sha256":"4f8af1456b6a805a470538bf6e0f82191774649c4ebd798748c8d01d7f021208"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:58.745281Z","signature_b64":"b+owQSDrya+Kj4cSIjQjqr8w5v21im2KhkURBMEF4Nwb7j+6jM/WQ45vtF2pGBuduukiJUDcbe5GT9n6SaD/AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4370a258a4a45b2002a4e02825cec36500c69acd3206c9239956417c60088bb0","last_reissued_at":"2026-05-26T01:03:58.744709Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:58.744709Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DUEL: Adversarial Self-Play for Multimodal Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bingjun Sun, Guangdeng Liao, Hanqing Zeng, Ji Liu, Lin Qiu, Yao Liu","submitted_at":"2026-05-24T00:38:26Z","abstract_excerpt":"Reinforcement learning (RL) has emerged as an effective paradigm for improving the reasoning capability of vision-language models (VLMs). However, RL-based optimization typically depends on costly high-quality annotations that are difficult to scale. Existing unsupervised alternatives may drift toward biased solutions due to weak visual grounding and the lack of reliable verification signals. We propose a self-evolving post-training framework, DUEL, where supervision emerges from adversarial interactions between two policies initialized from the same pretrained VLM. A Challenger generates an i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24794","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/2605.24794/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":"2605.24794","created_at":"2026-05-26T01:03:58.744783+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.24794v1","created_at":"2026-05-26T01:03:58.744783+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24794","created_at":"2026-05-26T01:03:58.744783+00:00"},{"alias_kind":"pith_short_12","alias_value":"INYKEWFEURNS","created_at":"2026-05-26T01:03:58.744783+00:00"},{"alias_kind":"pith_short_16","alias_value":"INYKEWFEURNSAAVE","created_at":"2026-05-26T01:03:58.744783+00:00"},{"alias_kind":"pith_short_8","alias_value":"INYKEWFE","created_at":"2026-05-26T01:03:58.744783+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/INYKEWFEURNSAAVE4AUCLTWDMU","json":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU.json","graph_json":"https://pith.science/api/pith-number/INYKEWFEURNSAAVE4AUCLTWDMU/graph.json","events_json":"https://pith.science/api/pith-number/INYKEWFEURNSAAVE4AUCLTWDMU/events.json","paper":"https://pith.science/paper/INYKEWFE"},"agent_actions":{"view_html":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU","download_json":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU.json","view_paper":"https://pith.science/paper/INYKEWFE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.24794&json=true","fetch_graph":"https://pith.science/api/pith-number/INYKEWFEURNSAAVE4AUCLTWDMU/graph.json","fetch_events":"https://pith.science/api/pith-number/INYKEWFEURNSAAVE4AUCLTWDMU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU/action/storage_attestation","attest_author":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU/action/author_attestation","sign_citation":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU/action/citation_signature","submit_replication":"https://pith.science/pith/INYKEWFEURNSAAVE4AUCLTWDMU/action/replication_record"}},"created_at":"2026-05-26T01:03:58.744783+00:00","updated_at":"2026-05-26T01:03:58.744783+00:00"}