{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:IGINEITWTVLJ477W3AH4PA3C4Y","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"c6d980674ff291a7786b841d8d358c9969a5678a4c379b3f084f249655e2ac0b","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-01-30T11:24:43Z","title_canon_sha256":"ce8c0a1f6a6cabc93275bdce39b18abc36a71a9a393efb409d87b8e1f81d4b47"},"schema_version":"1.0","source":{"id":"2601.22853","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2601.22853","created_at":"2026-05-18T02:45:05Z"},{"alias_kind":"arxiv_version","alias_value":"2601.22853v3","created_at":"2026-05-18T02:45:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.22853","created_at":"2026-05-18T02:45:05Z"},{"alias_kind":"pith_short_12","alias_value":"IGINEITWTVLJ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"IGINEITWTVLJ477W","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"IGINEITW","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:83121a97ad8f5c1e22127a54cbeba77a3615c870b5dde41f4a391491478cc76d","target":"graph","created_at":"2026-05-18T02:45:05Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"DyMo significantly outperforms state-of-the-art incomplete/dynamic MDL methods across various missing-data scenarios by adaptively identifying and fusing reliable recovered modalities using a task-loss proxy for information maximization."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The theoretically established connection between multimodal task-relevant information and the computable task loss at inference time is a valid and sufficient proxy for guiding modality selection without introducing bias from the recovery process."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DyMo dynamically selects reliable recovered modalities at inference by using task loss as a proxy for task-relevant information, outperforming prior discard-or-impute methods on image datasets."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"DyMo uses task loss computed at inference time as a proxy to select which recovered modalities to fuse for each sample."}],"snapshot_sha256":"4c878839df24453a9b4edc64c3d80947e7dc146e760c47249e86853a09d54d46"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Multimodal deep learning (MDL) has achieved remarkable success across various domains, yet its practical deployment is often hindered by incomplete multimodal data. Existing incomplete MDL methods either discard missing modalities, risking the loss of valuable task-relevant information, or recover them, potentially introducing irrelevant noise, leading to the discarding-imputation dilemma. To address this dilemma, in this paper, we propose DyMo, a new inference-time dynamic modality selection framework that adaptively identifies and fuses reliable recovered modalities, fully exploring task-rel","authors_text":"Chen Qin, Declan P. O'Regan, Siyi Du, Xinzhe Luo","cross_cats":[],"headline":"DyMo uses task loss computed at inference time as a proxy to select which recovered modalities to fuse for each sample.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-01-30T11:24:43Z","title":"Inference-Time Dynamic Modality Selection for Incomplete Multimodal Classification"},"references":{"count":13,"internal_anchors":2,"resolved_work":13,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Best of both worlds: Multimodal contrastive learning with tabular and imaging data","work_id":"a3119c36-85d6-4f78-a066-112f01bd8fad","year":2026},{"cited_arxiv_id":"1412.6980","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Multi- modal deep learning","work_id":"c4ab8fb8-7547-4e71-b74a-f3fd69725041","year":2026},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Deep learning and the information bottleneck principle","work_id":"2eb7426b-7351-4880-a100-59b960672668","year":2015},{"cited_arxiv_id":"2409.07825","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Deep Multimodal Learning with Missing Modality: A Survey","work_id":"ec0921fb-9035-4299-af61-70ccaf117c73","year":null}],"snapshot_sha256":"3de662a71b037422ec0dd35efc72174974047a6d00c0f668dcd5ffbb30fa76c5"},"source":{"id":"2601.22853","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-16T09:58:44.346585Z","id":"4169dffe-d6f3-4814-8513-bfc3ac9d96c2","model_set":{"reader":"grok-4.3"},"one_line_summary":"DyMo dynamically selects reliable recovered modalities at inference by using task loss as a proxy for task-relevant information, outperforming prior discard-or-impute methods on image datasets.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"DyMo uses task loss computed at inference time as a proxy to select which recovered modalities to fuse for each sample.","strongest_claim":"DyMo significantly outperforms state-of-the-art incomplete/dynamic MDL methods across various missing-data scenarios by adaptively identifying and fusing reliable recovered modalities using a task-loss proxy for information maximization.","weakest_assumption":"The theoretically established connection between multimodal task-relevant information and the computable task loss at inference time is a valid and sufficient proxy for guiding modality selection without introducing bias from the recovery process."}},"verdict_id":"4169dffe-d6f3-4814-8513-bfc3ac9d96c2"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:9f75868d097af568b42f8c6cf50e7c5ba6d7c1fe9569fed0014f8018387ff3ae","target":"record","created_at":"2026-05-18T02:45:05Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"c6d980674ff291a7786b841d8d358c9969a5678a4c379b3f084f249655e2ac0b","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-01-30T11:24:43Z","title_canon_sha256":"ce8c0a1f6a6cabc93275bdce39b18abc36a71a9a393efb409d87b8e1f81d4b47"},"schema_version":"1.0","source":{"id":"2601.22853","kind":"arxiv","version":3}},"canonical_sha256":"4190d222769d569e7ff6d80fc78362e60d465632a9b1e2a02679499525fd44ba","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4190d222769d569e7ff6d80fc78362e60d465632a9b1e2a02679499525fd44ba","first_computed_at":"2026-05-18T02:45:05.650612Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:45:05.650612Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"hvbYgEKO0aazor+fe2n6LvvVZUnT+PYzo9LRV0QNBiRGzg/R7JEnsX5UyKOJeLaANRz6YboIJ5BSirS52WfrBA==","signature_status":"signed_v1","signed_at":"2026-05-18T02:45:05.651307Z","signed_message":"canonical_sha256_bytes"},"source_id":"2601.22853","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9f75868d097af568b42f8c6cf50e7c5ba6d7c1fe9569fed0014f8018387ff3ae","sha256:83121a97ad8f5c1e22127a54cbeba77a3615c870b5dde41f4a391491478cc76d"],"state_sha256":"5e200b7a9b567b9dc16aed847db61d0d0d4b76d3334177abe99d1c2cc8683e5e"}