{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ENVO25VJU7RSQ667WMAT3IL5IS","short_pith_number":"pith:ENVO25VJ","schema_version":"1.0","canonical_sha256":"236aed76a9a7e3287bdfb3013da17d44863cb44ca5788d7ee397f3ea68538ea9","source":{"kind":"arxiv","id":"2605.20674","version":1},"attestation_state":"computed","paper":{"title":"Modular Multimodal Classification Without Fine-Tuning: A Simple Compositional Approach","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditya Mehrotra, Herman Bergstr\\\"om, Rahul G. Krishnan","submitted_at":"2026-05-20T03:43:19Z","abstract_excerpt":"We introduce CoMET, \\textit{\\textbf{C}omposing \\textbf{M}odality \\textbf{E}ncoders with \\textbf{T}abular foundation models}, a simple yet highly competitive method for multimodal classification: pass each modality through a frozen pre-trained backbone, compress the resulting embeddings with PCA, and concatenate as input into a Tabular Foundation Model (TFM) for prediction. We show that PCA alone suffices to act as an adaptor yielding strong, robust performance across modalities. When the \\texttt{CLS} tokens of the foundation model align poorly with downstream tasks, we propose \\textbf{PALPooli"},"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":"2605.20674","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T03:43:19Z","cross_cats_sorted":[],"title_canon_sha256":"835587c5051351d9a67b0ac2ea86343e46b54c110e5d26646670c641aa37b8f7","abstract_canon_sha256":"03de7df0138b1b3c7f2b0b86ba9e382eb783e6cec5c3bed81ad2eec65bb6020c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:48.645687Z","signature_b64":"+G1MLn/oKuC/vWQ16wqfuf7njH+pobqtui0MyvKj393o6zLepu8SGeZRWdMeCK3tnF40svhau1DPk+MefUmsCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"236aed76a9a7e3287bdfb3013da17d44863cb44ca5788d7ee397f3ea68538ea9","last_reissued_at":"2026-05-21T01:04:48.645069Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:48.645069Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Modular Multimodal Classification Without Fine-Tuning: A Simple Compositional Approach","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditya Mehrotra, Herman Bergstr\\\"om, Rahul G. Krishnan","submitted_at":"2026-05-20T03:43:19Z","abstract_excerpt":"We introduce CoMET, \\textit{\\textbf{C}omposing \\textbf{M}odality \\textbf{E}ncoders with \\textbf{T}abular foundation models}, a simple yet highly competitive method for multimodal classification: pass each modality through a frozen pre-trained backbone, compress the resulting embeddings with PCA, and concatenate as input into a Tabular Foundation Model (TFM) for prediction. We show that PCA alone suffices to act as an adaptor yielding strong, robust performance across modalities. When the \\texttt{CLS} tokens of the foundation model align poorly with downstream tasks, we propose \\textbf{PALPooli"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20674","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/2605.20674/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":"2605.20674","created_at":"2026-05-21T01:04:48.645174+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.20674v1","created_at":"2026-05-21T01:04:48.645174+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20674","created_at":"2026-05-21T01:04:48.645174+00:00"},{"alias_kind":"pith_short_12","alias_value":"ENVO25VJU7RS","created_at":"2026-05-21T01:04:48.645174+00:00"},{"alias_kind":"pith_short_16","alias_value":"ENVO25VJU7RSQ667","created_at":"2026-05-21T01:04:48.645174+00:00"},{"alias_kind":"pith_short_8","alias_value":"ENVO25VJ","created_at":"2026-05-21T01:04:48.645174+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/ENVO25VJU7RSQ667WMAT3IL5IS","json":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS.json","graph_json":"https://pith.science/api/pith-number/ENVO25VJU7RSQ667WMAT3IL5IS/graph.json","events_json":"https://pith.science/api/pith-number/ENVO25VJU7RSQ667WMAT3IL5IS/events.json","paper":"https://pith.science/paper/ENVO25VJ"},"agent_actions":{"view_html":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS","download_json":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS.json","view_paper":"https://pith.science/paper/ENVO25VJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.20674&json=true","fetch_graph":"https://pith.science/api/pith-number/ENVO25VJU7RSQ667WMAT3IL5IS/graph.json","fetch_events":"https://pith.science/api/pith-number/ENVO25VJU7RSQ667WMAT3IL5IS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS/action/storage_attestation","attest_author":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS/action/author_attestation","sign_citation":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS/action/citation_signature","submit_replication":"https://pith.science/pith/ENVO25VJU7RSQ667WMAT3IL5IS/action/replication_record"}},"created_at":"2026-05-21T01:04:48.645174+00:00","updated_at":"2026-05-21T01:04:48.645174+00:00"}