CT-FineBench: Diagnostic Fidelity Benchmark for AI-Generated CT Reports

CT-FineBench is a diagnostic fidelity benchmark designed for fine-grained evaluation of AI-generated Computed Tomography reports. The benchmark addresses a critical gap where traditional metrics like BLEU and ROUGE fail to capture granular diagnostic accuracy for clinical use. It was created by csyrf and is accepted by the ACL 2026 main conference.

Use Cases

• Benchmarking AI-generated CT report fidelity based on granular diagnostic accuracy metrics
• Evaluating clinical utility of NLP models for radiology beyond lexical overlap
• Developing new evaluation metrics for medical text generation based on diagnostic granularity

Strengths

• Specifically designed to address a critical gap in evaluating clinical text generation
• Accepted by the ACL 2026 main conference, indicating peer-reviewed validation

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 2026-04-30 03:39:36