Newest papers

APED-CXR: Adaptive Privileged Evidence Distillation for Chest X-Ray Grounding with Image-Only Inference

Md Shihabul Islam Shovo, Nishi Kanta Paul, Jamil Fayyad · 2026 · arXiv

DOI: 10.48550/arXiv.2610.29740 · open-access PDF

Spanish

Los modelos compactos de rayos X del pecho (CXR) que localizan los hallazgos se adaptan a la implementación, pero las anotaciones de la caja son limitadas y los modelos multimodales conscientes de los informes son costosos en inferencia. Introducimos APED-CXR, que destila la evidencia espacial informada por el informe de un profesor multimodal congelado a un estudiante con parámetros de 11.1M que predice enfermedades y cajas de mapas de evidencia compartidos y requiere la imagen sola en el momento de la prueba. Durante el entrenamiento, el profesor utiliza el radiógrafo, el informe y la clase consultada para evaluar las regiones propuestas de los mapas de evidencia actuales del estudiante, y sus puntuaciones de apoyo sirven como entrenamiento.

Machine translation (Meta M2M100); the English original follows.

Abstract

Compact chest X-ray (CXR) models that localize findings suit deployment, but box annotations are limited and report-aware multimodal models are expensive at inference. We introduce APED-CXR, which distills report-informed spatial evidence from a frozen multimodal teacher into a student with 11.1M parameters that predicts diseases and boxes from shared evidence maps and requires the image alone at test time. During training, the teacher uses the radiograph, report, and queried class to evaluate regions proposed from the student's current evidence maps, and its support scores serve as training t

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29740, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange