Summary of Harp: a Challenging Human-annotated Math Reasoning Benchmark, by Albert S. Yue et al.
HARP: A challenging human-annotated math reasoning benchmark
by Albert S. Yue, Lovish Madaan, Ted Moskovitz, DJ Strouse, Aaditya K. Singh
First submitted to arxiv on: 11 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The abstract presents the HARP dataset, a collection of 5,409 math reasoning problems from US national competitions. The dataset is designed to challenge frontier models in math-related tasks, with a focus on automatic checkable answers using libraries like SymPy. The authors report that even strong models struggle on the hardest problems, achieving average accuracy rates below 50%. The dataset also features multiple-choice options and ground-truth solutions for analysis. Evaluations of various frontier models demonstrate their tendency to scale inference-time compute for more difficult problems. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary HARP is a new math reasoning dataset with thousands of problems from US national competitions. It’s like a big test book, but instead of just answers, it also has multiple-choice options and two correct solutions for each problem. The goal is to make AI models better at math by giving them lots of practice on different types of problems. So far, even the smartest AI models aren’t great at solving these problems, especially the really hard ones. |
Keywords
» Artificial intelligence » Inference
