dorsal/arxiv
View SchemaReaMIL: Reasoning- and Evidence-Aware Multiple Instance Learning for Whole-Slide Histopathology
| Authors | Hyun Do Jung, Jungwon Choi, Hwiyoung Kim |
|---|---|
| Categories | |
| ArXiv ID | 2601.10073vv1 |
| URL | https://arxiv.org/abs/2601.10073 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
We introduce ReaMIL (Reasoning- and Evidence-Aware MIL), a multiple instance learning approach for whole-slide histopathology that adds a light selection head to a strong MIL backbone. The head produces soft per-tile gates and is trained with a budgeted-sufficiency objective: a hinge loss that enforces the true-class probability to be $\geq \tau$ using only the kept evidence, under a sparsity budget on the number of selected tiles. The budgeted-sufficiency objective yields small, spatially compact evidence sets without sacrificing baseline performance. Across TCGA-NSCLC (LUAD vs. LUSC), TCGA-BRCA (IDC vs. Others), and PANDA, ReaMIL matches or slightly improves baseline AUC and provides quantitative evidence-efficiency diagnostics. On NSCLC, it attains AUC 0.983 with a mean minimal sufficient K (MSK) $\approx 8.2$ tiles at $\tau = 0.90$ and AUKC $\approx 0.864$, showing that class confidence rises sharply and stabilizes once a small set of tiles is kept. The method requires no extra supervision, integrates seamlessly with standard MIL training, and naturally yields slide-level overlays. We report accuracy alongside MSK, AUKC, and contiguity for rigorous evaluation of model behavior on WSIs.
{
"annotation_id": "77b7ede2-4ade-4436-9fb0-c1fb7256b159",
"date_created": "2026-02-17T05:53:24.412000Z",
"date_modified": "2026-02-17T05:53:24.412000Z",
"file_hash": "56a0a55ca59d3c7b3b860d5cb5602162d7e19307bf1929f9b4a25aebfa331026",
"private": false,
"record": {
"abstract": "We introduce ReaMIL (Reasoning- and Evidence-Aware MIL), a multiple instance learning approach for whole-slide histopathology that adds a light selection head to a strong MIL backbone. The head produces soft per-tile gates and is trained with a budgeted-sufficiency objective: a hinge loss that enforces the true-class probability to be $\\geq \\tau$ using only the kept evidence, under a sparsity budget on the number of selected tiles. The budgeted-sufficiency objective yields small, spatially compact evidence sets without sacrificing baseline performance. Across TCGA-NSCLC (LUAD vs. LUSC), TCGA-BRCA (IDC vs. Others), and PANDA, ReaMIL matches or slightly improves baseline AUC and provides quantitative evidence-efficiency diagnostics. On NSCLC, it attains AUC 0.983 with a mean minimal sufficient K (MSK) $\\approx 8.2$ tiles at $\\tau = 0.90$ and AUKC $\\approx 0.864$, showing that class confidence rises sharply and stabilizes once a small set of tiles is kept. The method requires no extra supervision, integrates seamlessly with standard MIL training, and naturally yields slide-level overlays. We report accuracy alongside MSK, AUKC, and contiguity for rigorous evaluation of model behavior on WSIs.",
"arxiv_id": "2601.10073",
"authors": [
"Hyun Do Jung",
"Jungwon Choi",
"Hwiyoung Kim"
],
"categories": [
"cs.CV",
"cs.AI"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "ReaMIL: Reasoning- and Evidence-Aware Multiple Instance Learning for Whole-Slide Histopathology",
"url": "https://arxiv.org/abs/2601.10073",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "c0c81e8c-6286-476a-93d8-3277945e2daa",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}