{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PI7DO5NWRCMOEM745WDBN73K7Y","short_pith_number":"pith:PI7DO5NW","schema_version":"1.0","canonical_sha256":"7a3e3775b68898e233fced8616ff6afe29a87f6c6b5fc11c8108fda753e063fc","source":{"kind":"arxiv","id":"2511.15572","version":3},"attestation_state":"computed","paper":{"title":"From Per-Image Low-Rank to Encoding Mismatch: Rethinking Feature Distillation in Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"An encoding mismatch from per-image low-rank features and rotating dataset subspaces blocks feature-map distillation for compressing Vision Transformers.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bonan Xu, Huiyuan Tian, Shijian Li","submitted_at":"2025-11-19T16:03:21Z","abstract_excerpt":"Feature-map knowledge distillation (KD) transfers internal representations well between comparably sized Vision Transformers (ViTs), but it often fails in compression. We revisit this failure and uncover a paradox. Sample-wise SVD shows that each image is highly compressible, which seems to suggest that a narrow student with a linear projector should match the teacher \"in principle\". However, a dataset-level view contradicts this intuition: PCA shows that the teacher is a union of low-rank subspaces with significant subspace rotation across inputs. We further introduce token-level Spectral Ene"},"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":"2511.15572","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-11-19T16:03:21Z","cross_cats_sorted":[],"title_canon_sha256":"f1c33b217f031c57c0dbc57daea71266dd59ce186ae3ee4396e325d8ce3cd7cb","abstract_canon_sha256":"119b4c5982a6a6abd874ce894e60e69d93cad4c8d187812e7ea4e5ba9e01feea"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T02:05:08.394554Z","signature_b64":"f9FEdV6GGM1l0PNvn/uUo5g9qEWEFLB+TdFmKMP4/mDQoQz+ABpQfRwG20FK3pc6N/5S8mz+2Gif8kl6nGiNDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a3e3775b68898e233fced8616ff6afe29a87f6c6b5fc11c8108fda753e063fc","last_reissued_at":"2026-05-27T02:05:08.393777Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T02:05:08.393777Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Per-Image Low-Rank to Encoding Mismatch: Rethinking Feature Distillation in Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"An encoding mismatch from per-image low-rank features and rotating dataset subspaces blocks feature-map distillation for compressing Vision Transformers.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bonan Xu, Huiyuan Tian, Shijian Li","submitted_at":"2025-11-19T16:03:21Z","abstract_excerpt":"Feature-map knowledge distillation (KD) transfers internal representations well between comparably sized Vision Transformers (ViTs), but it often fails in compression. We revisit this failure and uncover a paradox. Sample-wise SVD shows that each image is highly compressible, which seems to suggest that a narrow student with a linear projector should match the teacher \"in principle\". However, a dataset-level view contradicts this intuition: PCA shows that the teacher is a union of low-rank subspaces with significant subspace rotation across inputs. We further introduce token-level Spectral Ene"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We refer to this combined phenomenon as an encoding mismatch. We propose two minimal remedies, Lift or WideLast... On ImageNet-1K, these fixes revive feature KD for ViT compression, improving DeiT-Tiny distilled from CaiT-S24 from 74.86% to 77.53%/78.23% top-1 accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the observed per-image low-rank structure, dataset-level subspace rotations, and token-level spectral energy patterns are the primary and causal reasons for feature-map KD failure in compression, rather than other factors such as optimization dynamics or capacity limits.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"The paper identifies an encoding mismatch in ViT feature distillation from per-image compressibility versus dataset subspace rotations and broad spectral energy patterns, proposing Lift and WideLast remedies that improve DeiT-Tiny accuracy from 74.86% to 77.53-78.23% on ImageNet-1K.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An encoding mismatch from per-image low-rank features and rotating dataset subspaces blocks feature-map distillation for compressing Vision Transformers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"131f4edf674c8156b3ccbdc07db66a596225e0a2b54b99c187b1abe371a97035"},"source":{"id":"2511.15572","kind":"arxiv","version":3},"verdict":{"id":"279a9c32-b2cf-4cb6-8506-5765e92914fc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T20:19:20.348005Z","strongest_claim":"We refer to this combined phenomenon as an encoding mismatch. We propose two minimal remedies, Lift or WideLast... On ImageNet-1K, these fixes revive feature KD for ViT compression, improving DeiT-Tiny distilled from CaiT-S24 from 74.86% to 77.53%/78.23% top-1 accuracy.","one_line_summary":"The paper identifies an encoding mismatch in ViT feature distillation from per-image compressibility versus dataset subspace rotations and broad spectral energy patterns, proposing Lift and WideLast remedies that improve DeiT-Tiny accuracy from 74.86% to 77.53-78.23% on ImageNet-1K.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the observed per-image low-rank structure, dataset-level subspace rotations, and token-level spectral energy patterns are the primary and causal reasons for feature-map KD failure in compression, rather than other factors such as optimization dynamics or capacity limits.","pith_extraction_headline":"An encoding mismatch from per-image low-rank features and rotating dataset subspaces blocks feature-map distillation for compressing Vision Transformers."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2511.15572/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":"2511.15572","created_at":"2026-05-27T02:05:08.393884+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.15572v3","created_at":"2026-05-27T02:05:08.393884+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.15572","created_at":"2026-05-27T02:05:08.393884+00:00"},{"alias_kind":"pith_short_12","alias_value":"PI7DO5NWRCMO","created_at":"2026-05-27T02:05:08.393884+00:00"},{"alias_kind":"pith_short_16","alias_value":"PI7DO5NWRCMOEM74","created_at":"2026-05-27T02:05:08.393884+00:00"},{"alias_kind":"pith_short_8","alias_value":"PI7DO5NW","created_at":"2026-05-27T02:05:08.393884+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.07191","citing_title":"Attention Transfer Is Not Universally Effective for Vision Transformers","ref_index":48,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y","json":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y.json","graph_json":"https://pith.science/api/pith-number/PI7DO5NWRCMOEM745WDBN73K7Y/graph.json","events_json":"https://pith.science/api/pith-number/PI7DO5NWRCMOEM745WDBN73K7Y/events.json","paper":"https://pith.science/paper/PI7DO5NW"},"agent_actions":{"view_html":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y","download_json":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y.json","view_paper":"https://pith.science/paper/PI7DO5NW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.15572&json=true","fetch_graph":"https://pith.science/api/pith-number/PI7DO5NWRCMOEM745WDBN73K7Y/graph.json","fetch_events":"https://pith.science/api/pith-number/PI7DO5NWRCMOEM745WDBN73K7Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y/action/storage_attestation","attest_author":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y/action/author_attestation","sign_citation":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y/action/citation_signature","submit_replication":"https://pith.science/pith/PI7DO5NWRCMOEM745WDBN73K7Y/action/replication_record"}},"created_at":"2026-05-27T02:05:08.393884+00:00","updated_at":"2026-05-27T02:05:08.393884+00:00"}