Summary of Quantemp: a Real-world Open-domain Benchmark For Fact-checking Numerical Claims, by Venktesh V et al.
QuanTemp: A real-world open-domain benchmark for fact-checking numerical claims
by Venktesh V, Abhijit Anand, Avishek Anand, Vinay Setty
First submitted to arxiv on: 25 Mar 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 This paper presents QuanTemp, a dataset designed to tackle the challenge of verifying real-world numerical claims. The focus is on complex claims lacking precise information, unlike existing works that primarily deal with synthetic claims. The authors release QuanTemp, a multi-domain dataset featuring fine-grained metadata and an evidence collection without leakage. The goal is to address the limitations of existing solutions for numerical claim verification. The paper evaluates and compares various approaches, including claim decomposition-based methods and numerical understanding-based models. Notably, the best baseline achieves a macro-F1 score of 58.32, highlighting the challenges in this task. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps keep the internet accurate by creating a special dataset to check true or false statements that have numbers in them. This is different from previous work which focused on simple Wikipedia claims. The new dataset has lots of examples and information about each claim, making it harder for computers to cheat. The authors test different ways computers can verify these claims and find the best method gets 58% correct. This shows how hard this task is. |
Keywords
» Artificial intelligence » F1 score
