Generative Evaluation of Complex Reasoning in Large Language Models
Haowei Lin, Xiangyu Wang, Ruilin Yan, Baizhou Huang, Haotian Ye, Jianhua Zhu, Zihao Wang, James Zou, Jianzhu Ma, Yitao Liang
2025-04-09
Summary
This paper talks about KUMO, a test system that checks if AI language models can actually solve new problems by thinking them through, instead of just remembering answers from their training data.
What's the problem?
Current AI tests get memorized by models over time, making it hard to know if they’re really thinking or just repeating stuff they’ve seen before.
What's the solution?
KUMO creates endless new puzzles and challenges on the fly, mixing AI creativity with logic engines to test how well models adapt to fresh problems they’ve never seen.
Why it matters?
This helps build AI that truly understands and solves real-world problems, like medical diagnosis or scientific research, instead of just faking it with memorized answers.
Abstract
With powerful large language models (LLMs) demonstrating superhuman reasoning capabilities, a critical question arises: Do LLMs genuinely reason, or do they merely recall answers from their extensive, web-scraped training datasets? Publicly released benchmarks inevitably become contaminated once incorporated into subsequent LLM training sets, undermining their reliability as faithful assessments. To address this, we introduce KUMO, a generative evaluation framework designed specifically for assessing reasoning in LLMs. KUMO synergistically combines LLMs with symbolic engines to dynamically produce diverse, multi-turn reasoning tasks that are partially observable and adjustable in difficulty. Through an automated pipeline, KUMO continuously generates novel tasks across open-ended domains, compelling models to demonstrate genuine generalization rather than memorization. We evaluated 23 state-of-the-art LLMs on 5,000 tasks across 100 domains created by KUMO, benchmarking their reasoning abilities against university students. Our findings reveal that many LLMs have outperformed university-level performance on easy reasoning tasks, and reasoning-scaled LLMs reach university-level performance on complex reasoning challenges. Moreover, LLM performance on KUMO tasks correlates strongly with results on newly released real-world reasoning benchmarks, underscoring KUMO's value as a robust, enduring assessment tool for genuine LLM reasoning capabilities.