\$OneMillion-Bench: How Far are Language Agents from Human Experts?

TL;DR

This paper introduces OneMillion-Bench, a comprehensive and challenging benchmark with 400 expert-curated tasks across multiple domains to evaluate language models' professional reasoning, source retrieval, and decision-making capabilities.

Contribution

It presents a new benchmark that emphasizes real-world professional tasks requiring complex reasoning, source validation, and domain-specific rules, filling gaps in existing evaluation methods.

Findings

Benchmark covers Law, Finance, Healthcare, and more.

Evaluates factual accuracy, coherence, and practical feasibility.

Provides a unified platform for assessing professional language model performance.

Abstract

As language models (LMs) evolve from chat assistants to long-horizon agents capable of multi-step reasoning and tool use, existing benchmarks remain largely confined to structured or exam-style tasks that fall short of real-world professional demands. To this end, we introduce \$OneMillion-Bench \$OneMillion-Bench, a benchmark of 400 expert-curated tasks spanning Law, Finance, Industry, Healthcare, and Natural Science, built to evaluate agents across economically consequential scenarios. Unlike prior work, the benchmark requires retrieving authoritative sources, resolving conflicting evidence, applying domain-specific rules, and making constraint decisions, where correctness depends as much on the reasoning process as the final answer. We adopt a rubric-based evaluation protocol scoring factual accuracy, logical coherence, practical feasibility, and professional compliance, focused on expert-level problems to ensure meaningful differentiation across agents. Together, \$OneMillion-Bench provides a unified testbed for assessing agentic reliability, professional depth, and practical readiness in domain-intensive scenarios.