{"paper":{"title":"Smoothing Slot Attention Iterations and Recurrences","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SmoothSA improves object-centric learning by smoothing slot attention iterations and recurrences.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Joni Pajarinen, Juho Kannala, Rongzhen Zhao, Wenyan Yang","submitted_at":"2025-08-07T14:09:33Z","abstract_excerpt":"Slot Attention (SA) lies at the heart of mainstream Object-Centric Learning (OCL). Image features can be aggregated into object-level representations by SA \\textit{iteratively} refining cold-start query slots. For video, such aggregation proceeds by SA \\textit{recurrently} shared across frames, with queries cold-started on the first frame while transitioned from the previous frame's slots thereafter. However, cold-start queries lack sample-specific cues thus hindering precise aggregation on image or video's first frame; Non-first frames' queries are already sample-specific thus requiring aggre"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We address these issues with our SmoothSA: (1) To smooth SA iterations on image or video's first frame, we preheat cold-start queries with rich input-feature information, by a tiny module self-distilled inside OCL; (2) To smooth SA recurrences across video's first and non-first frames, we differentiate the homogeneous aggregation transforms by using full and single iterations respectively.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that a self-distilled tiny module can reliably inject useful sample-specific cues into cold-start queries without introducing new biases or requiring extra supervision, and that switching between full and single iterations is sufficient to handle the information difference between first and non-first frames.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SmoothSA improves slot attention by preheating cold-start queries on first frames and applying full iterations there versus single iterations on subsequent video frames.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SmoothSA improves object-centric learning by smoothing slot attention iterations and recurrences.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ea8afb5ba87e619f22c3f43262e79208116852dd298a3a96d0619d5be91d1f46"},"source":{"id":"2508.05417","kind":"arxiv","version":5},"verdict":{"id":"cc500e45-6c67-46a0-b4c7-4e15c0e6fcbf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T23:55:55.082406Z","strongest_claim":"We address these issues with our SmoothSA: (1) To smooth SA iterations on image or video's first frame, we preheat cold-start queries with rich input-feature information, by a tiny module self-distilled inside OCL; (2) To smooth SA recurrences across video's first and non-first frames, we differentiate the homogeneous aggregation transforms by using full and single iterations respectively.","one_line_summary":"SmoothSA improves slot attention by preheating cold-start queries on first frames and applying full iterations there versus single iterations on subsequent video frames.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that a self-distilled tiny module can reliably inject useful sample-specific cues into cold-start queries without introducing new biases or requiring extra supervision, and that switching between full and single iterations is sufficient to handle the information difference between first and non-first frames.","pith_extraction_headline":"SmoothSA improves object-centric learning by smoothing slot attention iterations and recurrences."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.05417/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"}