[KongchangAI]
BenchmarkBRIGHT Benchmark

BRIGHT

An evaluation benchmark for reasoning-intensive retrieval that specifically assesses the performance of retrieval systems in scenarios requiring multi-step reasoning.

Timeline (last 90 days)

Oct 22

According to MarkTechPost, Rerank 5 Pro scored 82.2 on the BRIGHT benchmark.

Unverified50%
Oct 22

The BRIGHT benchmark specifically evaluates how retrieval systems perform in scenarios requiring multi-step reasoning, with typical tasks including math problem retrieval, code snippet matching and scientific reasoning document localization.

Unverified50%

All Facts (2)

Source Articles