{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5DRIVFIW7OTF6X34X6FVFJQRUH","short_pith_number":"pith:5DRIVFIW","schema_version":"1.0","canonical_sha256":"e8e28a9516fba65f5f7cbf8b52a611a1c09f9e88ccb3d55fa80b22fc10ce379a","source":{"kind":"arxiv","id":"2510.00705","version":3},"attestation_state":"computed","paper":{"title":"Training-free Uncertainty Guidance for Complex Visual Tasks with MLLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Rui Xiao, Sanghwan Kim, Stephan Alaniz, Yongqin Xian, Zeynep Akata","submitted_at":"2025-10-01T09:20:51Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) often struggle with fine-grained perception, such as identifying small objects in high-resolution images or detecting key moments in long videos. Existing methods typically rely on complex, task-specific fine-tuning, which reduces generalizability and increases system complexity. In this work, we propose an effective, training-free framework that uses an MLLM's intrinsic uncertainty as proactive guidance. Our core insight is that a model's uncertainty decreases when provided with relevant visual information. We introduce a unified mechanism that scores "},"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":"2510.00705","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-10-01T09:20:51Z","cross_cats_sorted":[],"title_canon_sha256":"f2b9be6ae67561b86ac2a9cd05da1b2c09a14c75074770c50c74b2ce0abf250e","abstract_canon_sha256":"2082b75952867c02bf4b6b1b22e0f39b2b28c39c9862b5deca4255ad8dd2a86d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-29T01:15:03.275089Z","signature_b64":"1Sepe4ImRIgWWiZHwUVuHBnE+7PAECbBsPDKLf/9AazepuWtiZwkrrlxqxZIPBCr5lUefpLslTmTymzPCgYCDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e8e28a9516fba65f5f7cbf8b52a611a1c09f9e88ccb3d55fa80b22fc10ce379a","last_reissued_at":"2026-06-29T01:15:03.274596Z","signature_status":"signed_v1","first_computed_at":"2026-06-29T01:15:03.274596Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training-free Uncertainty Guidance for Complex Visual Tasks with MLLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Rui Xiao, Sanghwan Kim, Stephan Alaniz, Yongqin Xian, Zeynep Akata","submitted_at":"2025-10-01T09:20:51Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) often struggle with fine-grained perception, such as identifying small objects in high-resolution images or detecting key moments in long videos. Existing methods typically rely on complex, task-specific fine-tuning, which reduces generalizability and increases system complexity. In this work, we propose an effective, training-free framework that uses an MLLM's intrinsic uncertainty as proactive guidance. Our core insight is that a model's uncertainty decreases when provided with relevant visual information. We introduce a unified mechanism that scores "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.00705","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/2510.00705/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":"2510.00705","created_at":"2026-06-29T01:15:03.274655+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.00705v3","created_at":"2026-06-29T01:15:03.274655+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.00705","created_at":"2026-06-29T01:15:03.274655+00:00"},{"alias_kind":"pith_short_12","alias_value":"5DRIVFIW7OTF","created_at":"2026-06-29T01:15:03.274655+00:00"},{"alias_kind":"pith_short_16","alias_value":"5DRIVFIW7OTF6X34","created_at":"2026-06-29T01:15:03.274655+00:00"},{"alias_kind":"pith_short_8","alias_value":"5DRIVFIW","created_at":"2026-06-29T01:15:03.274655+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.28421","citing_title":"DenoiseRL: Bootstrapping Reasoning Models to Recover from Noisy Prefixes","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2604.15376","citing_title":"Zoom Consistency: A Free Confidence Signal in Multi-Step Visual Grounding Pipelines","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH","json":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH.json","graph_json":"https://pith.science/api/pith-number/5DRIVFIW7OTF6X34X6FVFJQRUH/graph.json","events_json":"https://pith.science/api/pith-number/5DRIVFIW7OTF6X34X6FVFJQRUH/events.json","paper":"https://pith.science/paper/5DRIVFIW"},"agent_actions":{"view_html":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH","download_json":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH.json","view_paper":"https://pith.science/paper/5DRIVFIW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.00705&json=true","fetch_graph":"https://pith.science/api/pith-number/5DRIVFIW7OTF6X34X6FVFJQRUH/graph.json","fetch_events":"https://pith.science/api/pith-number/5DRIVFIW7OTF6X34X6FVFJQRUH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH/action/storage_attestation","attest_author":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH/action/author_attestation","sign_citation":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH/action/citation_signature","submit_replication":"https://pith.science/pith/5DRIVFIW7OTF6X34X6FVFJQRUH/action/replication_record"}},"created_at":"2026-06-29T01:15:03.274655+00:00","updated_at":"2026-06-29T01:15:03.274655+00:00"}