Benchmark RadarHAKARI-Bench
Evaluates retrieval architectures and efficiency settings (dimensionality reduction, quantization, reranking) across 35 benchmarks and 551 tasks in 43 languages, with unified conditions and metrics for BM25, dense, sparse, late interaction, and reranker models.
- Released
- 2026-06-22
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Fills the gap for a lightweight, high-fidelity proxy for full retrieval benchmarks, enabling rapid model selection, regression detection, and quality-efficiency trade-off analysis under consistent conditions, which is otherwise computationally prohibitive.
Motivation
With the rapid spread of retrieval-augmented generation and semantic search, choosing the right embedding and retrieval configuration is increasingly hard.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.