CiteVQA requires models to cite specific document regions with bounding boxes alongside answers and finds that even the strongest MLLMs frequently cite the wrong region, with top SAA scores of only 76.0 for closed models and 22.5 for open-source ones.
hub
What disease does this patient have? A large-scale open domain question answering dataset from medical exams
23 Pith papers cite this work. Polarity classification is still indexing.
23
Pith papers citing it
hub tools
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
CardioLens is a leakage-resistant CMR testbed of 473k slices and 13k QA pairs showing current MLLMs exhibit a large clinical reality gap with category-collapse failures on real workflows.
Checkup2Action is a new multimodal dataset and benchmark for generating safe, prioritized action cards from real-world clinical check-up reports using large language models.
A graphlet-anchored framework generates 119,856 factually grounded biomedical QA pairs that improve accuracy on PubMedQA and MedQA benchmarks.
SafeAnchor preserves 93.2% of original safety alignment across sequential domain adaptations by anchoring low-rank safety subspaces and constraining orthogonal updates, while matching unconstrained fine-tuning performance within 1.5 points.
Pico reduces LoRA merge interference by calibrating over-shared directions in the B matrix before merging, yielding 3.4-8.3 point accuracy gains and sometimes beating joint training.
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
DISeL augments standard LoRA with per-input gates over rank-one updates to reduce catastrophic forgetting during fine-tuning while adding few parameters.
OEP poisons self-evolving LLM agents by constructing clean edge-case experiences that appear locally valid yet cause harmful over-generalization during reflection, achieving over 50% attack success rate on GPT-4o agents across three domains.
CHI-Bench shows current AI agents achieve at most 28% success on long-horizon healthcare workflows that require dense policy adherence, multi-role handoffs, and multi-turn interactions.
MedExAgent models clinical diagnosis as a POMDP with patient and exam noise, then uses supervised fine-tuning followed by DAPO optimization to train an agent that matches larger models on diagnostic accuracy while controlling exam costs.
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
Exclusive Unlearning makes LLMs safe by forgetting all but retained domain knowledge, protecting against jailbreaks while preserving useful responses in areas like medicine and math.
MOSAIC is a training-free multi-agent LLM framework with rationale, coding, reflection, and debugging agents plus a consolidated context window that outperforms prior methods on scientific coding benchmarks.
SynthPert fine-tunes LLMs using synthetic reasoning traces to reach state-of-the-art on the PerturbQA benchmark for cellular perturbation prediction, surpassing the generating frontier model while generalizing to unseen cell types with only 2% of filtered data.
ArgLLMs build argumentation frameworks from LLMs to support explainable and contestable formal reasoning for claim verification.
C-MIG uses multi-view information gain from retrieved documents and refinements to supervise RAG-RL for clinical diagnosis, claiming top performance on four medical benchmarks.
MedFabric dataset and EtHER detector achieve over 15% better word-level fabrication detection in medical LLMs than prior methods by generating stylistically faithful errors and using decomposition-based checking.
Dual-Stream Calibration uses entropy minimization and iterative meta-learning at test time to internalize clinical evidence and outperform standard in-context learning baselines on medical tasks.
HeteroRAG integrates modality-specific retrieval from medical reports and multi-corpus text sources with preference tuning to improve factual accuracy in Med-LVLMs across 11 datasets.
Tag-based few-shot selection yields higher precision and stability than random or similarity-based methods when using LLMs to analyze medical incidents.
Dense retrieval plus query reformulation and reranking reaches 60.49% accuracy on MedQA USMLE, outperforming other setups while domain-specialized models make better use of the retrieved evidence.
citing papers explorer
-
CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence
CiteVQA requires models to cite specific document regions with bounding boxes alongside answers and finds that even the strongest MLLMs frequently cite the wrong region, with top SAA scores of only 76.0 for closed models and 22.5 for open-source ones.
-
CardioLens: Revealing the Clinical Reality Gap of MLLMs via Multi-Sequence Cardiac MRI Evaluations
CardioLens is a leakage-resistant CMR testbed of 473k slices and 13k QA pairs showing current MLLMs exhibit a large clinical reality gap with category-collapse failures on real workflows.
-
Checkup2Action: A Multimodal Clinical Check-up Report Dataset for Patient-Oriented Action Card Generation
Checkup2Action is a new multimodal dataset and benchmark for generating safe, prioritized action cards from real-world clinical check-up reports using large language models.
-
BioGraphletQA: Knowledge-Anchored Generation of Complex QA Datasets
A graphlet-anchored framework generates 119,856 factually grounded biomedical QA pairs that improve accuracy on PubMedQA and MedQA benchmarks.
-
SafeAnchor: Preventing Cumulative Safety Erosion in Continual Domain Adaptation of Large Language Models
SafeAnchor preserves 93.2% of original safety alignment across sequential domain adaptations by anchoring low-rank safety subspaces and constraining orthogonal updates, while matching unconstrained fine-tuning performance within 1.5 points.
-
Crowded in B-Space: Calibrating Shared Directions for LoRA Merging
Pico reduces LoRA merge interference by calibrating over-shared directions in the B matrix before merging, yielding 3.4-8.3 point accuracy gains and sometimes beating joint training.
-
PRIMETIME : Limits of LLMs in Temporal Primitives
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
-
Learning When to Adapt
DISeL augments standard LoRA with per-input gates over rank-one updates to reduce catastrophic forgetting during fine-tuning while adding few parameters.
-
OEP: Poisoning Self-Evolving LLM Agents via Locally Correct but Non-Transferable Experiences
OEP poisons self-evolving LLM agents by constructing clean edge-case experiences that appear locally valid yet cause harmful over-generalization during reflection, achieving over 50% attack success rate on GPT-4o agents across three domains.
-
CHI-Bench: Can AI Agents Automate End-to-End, Long-Horizon, Policy-Rich Healthcare Workflows?
CHI-Bench shows current AI agents achieve at most 28% success on long-horizon healthcare workflows that require dense policy adherence, multi-role handoffs, and multi-turn interactions.
-
MedExAgent: Training LLM Agents to Ask, Examine, and Diagnose in Noisy Clinical Environments
MedExAgent models clinical diagnosis as a POMDP with patient and exam noise, then uses supervised fine-tuning followed by DAPO optimization to train an agent that matches larger models on diagnostic accuracy while controlling exam costs.
-
Compared to What? Baselines and Metrics for Counterfactual Prompting
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
-
Exclusive Unlearning
Exclusive Unlearning makes LLMs safe by forgetting all but retained domain knowledge, protecting against jailbreaks while preserving useful responses in areas like medicine and math.
-
MOSAIC: Multi-agent Orchestration for Task-Intelligent Scientific Coding
MOSAIC is a training-free multi-agent LLM framework with rationale, coding, reflection, and debugging agents plus a consolidated context window that outperforms prior methods on scientific coding benchmarks.
-
SynthPert: Enhancing LLM Biological Reasoning via Synthetic Reasoning Traces for Cellular Perturbation Prediction
SynthPert fine-tunes LLMs using synthetic reasoning traces to reach state-of-the-art on the PerturbQA benchmark for cellular perturbation prediction, surpassing the generating frontier model while generalizing to unseen cell types with only 2% of filtered data.
-
Argumentative Large Language Models for Explainable and Contestable Claim Verification
ArgLLMs build argumentation frameworks from LLMs to support explainable and contestable formal reasoning for claim verification.
-
C-MIG: Multi-view Information Gain-based Retrieval-Augmented Generation for Clinical Diagnosis Reasoning
C-MIG uses multi-view information gain from retrieved documents and refinements to supervise RAG-RL for clinical diagnosis, claiming top performance on four medical benchmarks.
-
MedFabric and EtHER: A Data-Centric Framework for Word-Level Fabrication Generation and Detection in Medical LLMs
MedFabric dataset and EtHER detector achieve over 15% better word-level fabrication detection in medical LLMs than prior methods by generating stylistically faithful errors and using decomposition-based checking.
-
From Exposure to Internalization: Dual-Stream Calibration for In-context Clinical Reasoning
Dual-Stream Calibration uses entropy minimization and iterative meta-learning at test time to internalize clinical evidence and outperform standard in-context learning baselines on medical tasks.
-
HeteroRAG: A Heterogeneous Retrieval-Augmented Generation Framework for Medical Vision Language Tasks
HeteroRAG integrates modality-specific retrieval from medical reports and multi-corpus text sources with preference tuning to improve factual accuracy in Med-LVLMs across 11 datasets.
-
Medical Incident Causal Factors and Preventive Measures Generation Using Tag-based Example Selection in Few-shot Learning
Tag-based few-shot selection yields higher precision and stability than random or similarity-based methods when using LLMs to analyze medical incidents.
-
A Systematic Study of Retrieval Pipeline Design for Retrieval-Augmented Medical Question Answering
Dense retrieval plus query reformulation and reranking reaches 60.49% accuracy on MedQA USMLE, outperforming other setups while domain-specialized models make better use of the retrieved evidence.
- Query-efficient model evaluation using cached responses