{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HB4I3UALQBOZ3ZS7IIR3XBQZNP","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":"0b9a6b2ddb14a5553b9f297063d56b467b5afa596ab3f582a4f4dba2f0101d7b","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"eess.IV","submitted_at":"2026-05-12T12:15:41Z","title_canon_sha256":"4bebf2b53ff5d0b3b0a0a21b74c1bae8ab81291330f12c75ddaa146124f373b4"},"schema_version":"1.0","source":{"id":"2605.12575","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12575","created_at":"2026-05-18T03:10:01Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12575v1","created_at":"2026-05-18T03:10:01Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12575","created_at":"2026-05-18T03:10:01Z"},{"alias_kind":"pith_short_12","alias_value":"HB4I3UALQBOZ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HB4I3UALQBOZ3ZS7","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HB4I3UAL","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:4e173d5e0d57cf81884325f7d03865b171013fa22f178959db01ca0e59181512","target":"graph","created_at":"2026-05-18T03:10:01Z","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":"Across three WSI benchmarks and seven MIL backbones, FOCI reveals that compact rationales are selection-headroom dependent: transformer and multi-branch attention aggregators can admit compact rationales, near-minimal attention-pooling baselines enter a selection-saturation regime, and hard-selection backbones can conflict with an external readout. For TransMIL, relative to its documented CLS-proxy ranking, FOCI reduces the Minimum Sufficient K (MSK) tile count by 32-56% across benchmarks, while ACMIL+FOCI attains the highest mean SHI (+0.465)."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the sufficiency and exclusion objectives used to train FOCI produce tile subsets that are genuinely model-sufficient and free of readout-induced artifacts, and that the Sequential Reveal Protocol accurately isolates selection headroom without confounding effects from the specific keep/drop training procedure."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"FOCI adds a post-hoc readout to frozen WSI-MIL models to find compact output-consistent tile subsets and measures selection headroom with SHI, showing transformer-based models allow smaller rationales than attention-pooling baselines."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"FOCI reveals that compact rationales for frozen WSI-MIL predictions depend on the choice of backbone aggregator."}],"snapshot_sha256":"56502401941b1f087f2f46c46f7a8a30549c1e1272395da7c0690c2672dd5952"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Whole-slide image (WSI) multiple instance learning (MIL) classifiers can achieve strong slide-level AUC while leaving the full-bag prediction opaque. Attention scores are widely reused as post-hoc explanations, but high attention can reflect aggregation preference rather than a compact, model-sufficient rationale. We study post-hoc rationale highlighting for frozen WSI-MIL: given a trained classifier, can its slide-level prediction be recovered from a compact, output-consistent tile subset without retraining the backbone? We instantiate this with Finding Optimal Contextual Instances (FOCI), a ","authors_text":"Hwiyoung Kim, Hyun Do Jung, Jungwon Choi, Soojung Choi, Yujin Oh","cross_cats":["cs.AI","cs.CV"],"headline":"FOCI reveals that compact rationales for frozen WSI-MIL predictions depend on the choice of backbone aggregator.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"eess.IV","submitted_at":"2026-05-12T12:15:41Z","title":"Are Compact Rationales Free? Measuring Tile Selection Headroom in Frozen WSI-MIL"},"references":{"count":41,"internal_anchors":0,"resolved_work":41,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Hanna, Luke Geneslaw, Allen Miraflor, Vitor Werneck Krauss Silva, Klaus J","work_id":"9c9d09a0-b465-413e-b9f4-4b6815b21e87","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Attention-based deep multiple instance learning","work_id":"8ef34d56-4419-46b3-ba9c-ec64c047299a","year":2018},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Data-efficient and weakly supervised computational pathology on whole-slide images.Nature biomedical engineering, 5(6):555–570","work_id":"d10a268e-9baa-47ed-bd74-87bb2719f105","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Towards a general-purpose foundation model for computational pathology.Nature medicine, 30(3):850–862","work_id":"3630134d-7234-4fdb-a23f-689393e31801","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Deep learning for whole slide image analysis: an overview.Frontiers in medicine, 6:264, 2019","work_id":"2a5e3d11-c406-47b5-9eb8-ed322351536f","year":2019}],"snapshot_sha256":"89e2d24ff836abadcd54d43675b028df970358857dd063067defab1b13bf0d56"},"source":{"id":"2605.12575","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T20:35:07.746936Z","id":"d0d55ff1-8d7f-4010-90db-c7f5c4cef935","model_set":{"reader":"grok-4.3"},"one_line_summary":"FOCI adds a post-hoc readout to frozen WSI-MIL models to find compact output-consistent tile subsets and measures selection headroom with SHI, showing transformer-based models allow smaller rationales than attention-pooling baselines.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"FOCI reveals that compact rationales for frozen WSI-MIL predictions depend on the choice of backbone aggregator.","strongest_claim":"Across three WSI benchmarks and seven MIL backbones, FOCI reveals that compact rationales are selection-headroom dependent: transformer and multi-branch attention aggregators can admit compact rationales, near-minimal attention-pooling baselines enter a selection-saturation regime, and hard-selection backbones can conflict with an external readout. For TransMIL, relative to its documented CLS-proxy ranking, FOCI reduces the Minimum Sufficient K (MSK) tile count by 32-56% across benchmarks, while ACMIL+FOCI attains the highest mean SHI (+0.465).","weakest_assumption":"That the sufficiency and exclusion objectives used to train FOCI produce tile subsets that are genuinely model-sufficient and free of readout-induced artifacts, and that the Sequential Reveal Protocol accurately isolates selection headroom without confounding effects from the specific keep/drop training procedure."}},"verdict_id":"d0d55ff1-8d7f-4010-90db-c7f5c4cef935"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:49300a512d775d5d4cd9ee94c1c7362021603a9d01eff123769693130903a558","target":"record","created_at":"2026-05-18T03:10:01Z","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":"0b9a6b2ddb14a5553b9f297063d56b467b5afa596ab3f582a4f4dba2f0101d7b","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"eess.IV","submitted_at":"2026-05-12T12:15:41Z","title_canon_sha256":"4bebf2b53ff5d0b3b0a0a21b74c1bae8ab81291330f12c75ddaa146124f373b4"},"schema_version":"1.0","source":{"id":"2605.12575","kind":"arxiv","version":1}},"canonical_sha256":"38788dd00b805d9de65f4223bb86196be2dd3c452bdb550061b9411405d22132","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"38788dd00b805d9de65f4223bb86196be2dd3c452bdb550061b9411405d22132","first_computed_at":"2026-05-18T03:10:01.648828Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:10:01.648828Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZU7WW/K2XpuGfPZfS+iui1nAY7GN7nSf7lDQKNyHGxkSdnzzn1Zj23lFDyJTUbLtKxsZwjRleCyzNIWc1fGABA==","signature_status":"signed_v1","signed_at":"2026-05-18T03:10:01.649723Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.12575","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:49300a512d775d5d4cd9ee94c1c7362021603a9d01eff123769693130903a558","sha256:4e173d5e0d57cf81884325f7d03865b171013fa22f178959db01ca0e59181512"],"state_sha256":"7f92a9bdcbaf5483e90ab772f9d98c4b1bf73ceaca9b49ef8d5c8a0f28bdf5a2"}