MTR-Suite: A Framework for Evaluating and Synthesizing Conversational Retrieval Benchmarks

In three linesMTR-Suite is a framework for auditing and synthesizing conversational retrieval benchmarks. It includes MTR-Eval (LLM-based auditor), MTR-Pipeline (multi-agent system generating high-fidelity dialogues at 1/400th human cost), and MTR-Bench (general-domain benchmark). Code and data publicly available.

Summary generated by Claude — human-verified