Have LLMs Advanced Enough? A Challenging Problem Solving Benchmark For Large Language Models

TL;DR

This paper introduces JEEBench, a highly challenging problem-solving benchmark for large language models, highlighting current limitations and proposing a confidence-based method to improve response accuracy.

Contribution

The paper presents JEEBench, a new difficult benchmark with 515 complex problems, and develops a confidence-thresholding technique to enhance model response selection.

Findings

Even the best models perform below 40% accuracy.

Common failure modes include algebra errors and concept grounding issues.

Prompting alone cannot effectively assess answer risk.

Abstract

The performance of large language models (LLMs) on existing reasoning benchmarks has significantly improved over the past years. In response, we present JEEBench, a considerably more challenging benchmark dataset for evaluating the problem solving abilities of LLMs. We curate 515 challenging pre-engineering mathematics, physics and chemistry problems from the highly competitive IIT JEE-Advanced exam. Long-horizon reasoning on top of deep in-domain knowledge is essential for solving problems in this benchmark. Our evaluation on various open-source and proprietary models reveals that the highest performance, even after using techniques like self-consistency, self-refinement and chain-of-thought prompting, is less than 40%. The typical failure modes of GPT-4, the best model, are errors in algebraic manipulation, difficulty in grounding abstract concepts into mathematical equations accurately and failure in retrieving relevant domain-specific concepts. We also observe that by mere prompting, GPT-4 is unable to assess risk introduced by negative marking for incorrect answers. For this, we develop a post-hoc confidence-thresholding method over self-consistency, which enables effective response selection. We hope that our challenging benchmark will guide future re-search in problem-solving using LLMs.