MTBBench: A Multimodal Clinical Decision-Making Benchmark in Oncology

MTBBench is a benchmark designed to evaluate multimodal large language models in complex clinical oncology scenarios. It focuses on multimodal integration of data like pathology and radiology, and longitudinal reasoning across patient timelines. The benchmark was authored by EeshaanJain and last updated on October 23, 2025.

Use Cases

• Benchmarking multimodal AI reasoning based on pathology, genomics, and radiology data mentioned in the description
• Evaluating longitudinal reasoning capabilities based on sequential patient timelines
• Testing agentic AI interaction tasks based on the requirement to interface with external tools

Strengths

• Focuses on two core challenges in oncology: multimodal integration and longitudinal reasoning
• Designed for agentic tasks requiring interaction with external tools

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count and dataset size are unknown, which may limit suitability assessment

Provenance

Source

huggingface

Freshness

Last updated 2025-10-23 09:23:50