VibeSearchBench is a dataset containing 200 simulated research tasks for evaluating search systems. It includes 100 daily-life tasks and 100 professional tasks, each with a ground-truth knowledge graph. The dataset was created by VibeSearchBench and was last updated on May 6, 2026.
Use Cases
- Benchmark search system performance based on simulated daily-life and professional research tasks.
- Evaluate knowledge graph construction or retrieval algorithms based on provided ground-truth nodes and triples.
- Test multilingual search capabilities based on the presence of English (en) and Chinese (zh) language fields.
- Simulate personalized search interactions based on the provided user persona field.
Strengths
- Contains 200 distinct tasks, split evenly between daily-life and professional contexts.
- Provides ground-truth knowledge graph structures (nodes and triples) for each task.
- Includes multilingual content with English and Chinese language tags.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- VibeSearchBench
- Freshness
- Last updated 2026-05-06 15:28:22; freshness should be verified.