Humanity's Last Exam
cs.LG, cs.AI, cs.CL
Submitted: 2025-01-24
Updated: 2026-10-19
Comments: 29 pages, 6 figures
DOI: 10.1038/s41586-025-09962-4
Code: https://github.com/deepseek-ai/DeepSeek-V3
License: http://creativecommons.org/licenses/by/4.0/
The gist: Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities.
Terminology
Abstract
Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.
Sources
- A BERT Baseline for the Natural Questions
- AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents
- HealthBench: Evaluating Large Language Models Towards Improved Human Health
- Program Synthesis with Large Language Models
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Purple Llama CyberSecEval: A Secure Coding Benchmark for Language Models
- MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering
- Evaluating Large Language Models Trained on Code
- ARC Prize 2024: Technical Report
- Training Verifiers to Solve Math Word Problems
- DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs
- Omni-MATH: A Universal Olympiad Level Mathematic Benchmark For Large Language Models
- FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI
- Measuring Coding Challenge Competence With APPS
- Not All LLM Reasoners Are Created Equal
- Refusal-Trained LLMs Are Easily Jailbroken As Browser Agents
- LAB-Bench: Measuring Capabilities of Language Models for Biology Research
- The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning
- MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts
- Inadequacies of Large Language Model Benchmarks in the Era of Generative Artificial Intelligence
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks