Benchmark RadarAI BENCHMARK PROFILE
CADP-Bench
CADP-Bench evaluates multimodal LLMs on compilable academic document parsing. It includes expert-verified full academic pages with tightly coupled text and structured elements, assessed through a re-injection compilation protocol.
- Released
- 2026-08-18
- Readiness
- Runnable
- Primary field
- Science & Research
Why it matters
Provides a structured evaluation for structure-aware scientific document parsing, measuring the fidelity of executable reconstructions, which is crucial for machine-readable scientific knowledge.
Motivation
Academic papers are a primary carrier of scientific knowledge, yet most of this knowledge remains locked in PDFs that are optimized for human reading rather than machine use.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.