Benchmark RadarDailyReport
DailyReport is a benchmark with 150 open-ended daily search tasks and 3,546 rubrics, evaluating search agents on information-seeking tasks. Tasks are decomposed into subtasks with cascade rubrics across dimensions, yielding interpretable scores and a user preference score.
- Released
- 2026-06-11
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing search agent benchmarks often use specialized or artificial tasks with coarse scoring, limiting interpretability and real-world relevance. DailyReport provides a more user-centric evaluation protocol for daily search tasks, offering fine-grained, dimension-level scores to inform agent development.
Motivation
Search Agents (SAs) typically leverage large language models (LLMs) to support complex information-seeking tasks by autonomously exploring web sources and synthesizing information into comprehensive responses.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.