{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CYHI64ITHHWNTOJKZJIJGWLWXU","short_pith_number":"pith:CYHI64IT","schema_version":"1.0","canonical_sha256":"160e8f711339ecd9b92aca50935976bd0ebcfa9925cc0fd95abf32a0af68282a","source":{"kind":"arxiv","id":"2506.15649","version":1},"attestation_state":"computed","paper":{"title":"Dual-Stage Value-Guided Inference with Margin-Based Reward Adjustment for Fast and Faithful VLM Captioning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adinath Madhavrao Dukre, Ankan Deria, Feilong Tang, Imran Razzak, Muhammad Awais, Muhammad Haris Khan, Sara Atito, Sudipta Roy","submitted_at":"2025-06-18T17:23:36Z","abstract_excerpt":"Despite significant advances in inference-time search for vision-language models (VLMs), existing approaches remain both computationally expensive and prone to unpenalized, low-confidence generations which often lead to persistent hallucinations. We introduce \\textbf{Value-guided Inference with Margin-based Reward (ViMaR)}, a two-stage inference framework that improves both efficiency and output fidelity by combining a temporal-difference value model with a margin-aware reward adjustment. In the first stage, we perform a single pass to identify the highest-value caption among diverse candidate"},"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":"2506.15649","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-06-18T17:23:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f8e709b00ec8ca793c8affb4bec9b0ebb9cc468c1f5055d7f582621d076b199e","abstract_canon_sha256":"663be99f3525a115182866bec72777b5e38161e7db1073b51bc4b2a7af9cab44"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:23:47.771684Z","signature_b64":"ef1fidor4kAl9Pbo7hn4N2ujYQvtdZyGspJBj1kCFuKoEa6GuAkVRnHauDDoARGuBT3yfWLpiznfHM2AhSFlDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"160e8f711339ecd9b92aca50935976bd0ebcfa9925cc0fd95abf32a0af68282a","last_reissued_at":"2026-07-05T11:23:47.770929Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:23:47.770929Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dual-Stage Value-Guided Inference with Margin-Based Reward Adjustment for Fast and Faithful VLM Captioning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Adinath Madhavrao Dukre, Ankan Deria, Feilong Tang, Imran Razzak, Muhammad Awais, Muhammad Haris Khan, Sara Atito, Sudipta Roy","submitted_at":"2025-06-18T17:23:36Z","abstract_excerpt":"Despite significant advances in inference-time search for vision-language models (VLMs), existing approaches remain both computationally expensive and prone to unpenalized, low-confidence generations which often lead to persistent hallucinations. We introduce \\textbf{Value-guided Inference with Margin-based Reward (ViMaR)}, a two-stage inference framework that improves both efficiency and output fidelity by combining a temporal-difference value model with a margin-aware reward adjustment. In the first stage, we perform a single pass to identify the highest-value caption among diverse candidate"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.15649","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/2506.15649/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":"2506.15649","created_at":"2026-07-05T11:23:47.771019+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15649v1","created_at":"2026-07-05T11:23:47.771019+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15649","created_at":"2026-07-05T11:23:47.771019+00:00"},{"alias_kind":"pith_short_12","alias_value":"CYHI64ITHHWN","created_at":"2026-07-05T11:23:47.771019+00:00"},{"alias_kind":"pith_short_16","alias_value":"CYHI64ITHHWNTOJK","created_at":"2026-07-05T11:23:47.771019+00:00"},{"alias_kind":"pith_short_8","alias_value":"CYHI64IT","created_at":"2026-07-05T11:23:47.771019+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.09749","citing_title":"See Fair, Speak Truth: Equitable Attention Improves Grounding and Reduces Hallucination in Vision-Language Alignment","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU","json":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU.json","graph_json":"https://pith.science/api/pith-number/CYHI64ITHHWNTOJKZJIJGWLWXU/graph.json","events_json":"https://pith.science/api/pith-number/CYHI64ITHHWNTOJKZJIJGWLWXU/events.json","paper":"https://pith.science/paper/CYHI64IT"},"agent_actions":{"view_html":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU","download_json":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU.json","view_paper":"https://pith.science/paper/CYHI64IT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15649&json=true","fetch_graph":"https://pith.science/api/pith-number/CYHI64ITHHWNTOJKZJIJGWLWXU/graph.json","fetch_events":"https://pith.science/api/pith-number/CYHI64ITHHWNTOJKZJIJGWLWXU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU/action/storage_attestation","attest_author":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU/action/author_attestation","sign_citation":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU/action/citation_signature","submit_replication":"https://pith.science/pith/CYHI64ITHHWNTOJKZJIJGWLWXU/action/replication_record"}},"created_at":"2026-07-05T11:23:47.771019+00:00","updated_at":"2026-07-05T11:23:47.771019+00:00"}