Summary of Mathador-lm: a Dynamic Benchmark For Mathematical Reasoning on Large Language Models, by Eldar Kurtic et al.
Mathador-LM: A Dynamic Benchmark for Mathematical Reasoning on Large Language Models
by Eldar Kurtic, Amir Moeini, Dan Alistarh
First submitted to arxiv on: 18 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 A new benchmark, Mathador-LM, is introduced for evaluating the mathematical reasoning capabilities of large language models (LLMs). This benchmark combines ruleset interpretation, planning, and problem-solving to assess a model’s ability to solve math problems. The benchmark is inspired by the Mathador game, where users must reach a target number using basic arithmetic operations on a given set of base numbers. The authors show that leading LLMs achieve stable average performance while generating benchmark instances dynamically, following a target difficulty level. This alleviates concerns about test-set leakage into training data, which often undermines popular benchmarks. The authors also conduct an evaluation of state-of-the-art LLMs on Mathador-LM and find that contemporary models struggle with the task, scoring lower than average 3rd graders. In contrast, they perform well on popular mathematical reasoning benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Mathador-LM is a new test for big language models to see if they can do math problems. The model has to follow rules and solve problems to get to a target number using basic addition and subtraction. This helps make sure the tests aren’t too easy or too hard, which is a problem with some other math tests. Researchers tested many different language models and found that most of them struggle to do well on this test, even though they can do more advanced math problems. |
