Rethinking Reasoning-Intensive Retrieval: Evaluating and Advancing Retrievers in Agentic Search Systems

TL;DR

This paper introduces a new benchmark and training corpus for reasoning-intensive retrieval, emphasizing evidence diversity and agentic search, and demonstrates improved retriever performance through aspect-aware evaluation and fine-tuning.

Contribution

It presents BRIGHT-Pro, an expanded benchmark with multi-aspect evidence, and RTriever-Synth, a synthetic corpus for training, advancing retrieval evaluation and training methods.

Findings

RTriever-4B outperforms its base model in experiments.

Aspect-aware evaluation reveals behaviors hidden by standard metrics.

BRIGHT-Pro provides multi-aspect gold evidence for better assessment.

Abstract

Reasoning-intensive retrieval aims to surface evidence that supports downstream reasoning rather than merely matching topical similarity. This capability is increasingly important for agentic search systems, where retrievers must provide complementary evidence across iterative search and synthesis. However, existing work remains limited on both evaluation and training: benchmarks such as BRIGHT provide narrow gold sets and evaluate retrievers in isolation, while synthetic training corpora often optimize single-passage relevance rather than evidence portfolio construction. We introduce BRIGHT-Pro, an expert-annotated benchmark that expands each query with multi-aspect gold evidence and evaluates retrievers under both static and agentic search protocols. We further construct RTriever-Synth, an aspect-decomposed synthetic corpus that generates complementary positives and positive-conditioned hard negatives, and use it to LoRA fine-tune RTriever-4B from Qwen3-Embedding-4B. Experiments across lexical, general-purpose, and reasoning-intensive retrievers show that aspect-aware and agentic evaluation expose behaviors hidden by standard metrics, while RTriever-4B substantially improves over its base model.