Benchmark RadarAI BENCHMARK PROFILE
ARAC-Bench
ARAC-Bench evaluates auto-research systems on alignment and completeness of research trajectories across proposal, experiment, and synthesis stages using rubrics derived from academic cognition skills.
- Released
- 2026-08-13
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Provides a standardized framework for assessing autonomous research systems beyond final output, enabling comparison of research process quality.
Motivation
The rapid advancement of Auto-Research has surfaced a fundamental evaluation challenge: how can we measure the alignment, logical coherence, and evolutionary completeness of its research trajectory with human research behavior?
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.