Summary of Perteval: Unveiling Real Knowledge Capacity Of Llms with Knowledge-invariant Perturbations, by Jiatong Li et al.
PertEval: Unveiling Real Knowledge Capacity of LLMs with Knowledge-Invariant Perturbations
by Jiatong Li, Renjun Hu, Kunzhe Huang, Yan Zhuang, Qi Liu, Mengxiao Zhu, Xing Shi, Wei Lin
First submitted to arxiv on: 30 May 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
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 paper proposes a toolkit called PertEval for assessing the knowledge capacity of large language models (LLMs) through in-depth probing. It addresses concerns about the reliability of existing benchmarks by employing human-like restatement techniques to generate test samples from static benchmarks while retaining knowledge-critical content. The toolkit includes response consistency analyses to compare performance on raw and perturbed test sets, providing a more accurate assessment of LLMs’ genuine knowledge capacity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary PertEval is a new way to test how well large language models (LLMs) really know things. Right now, we use tests that are designed by experts, but these tests might not be very good because they’re limited and can get contaminated with extra information. PertEval changes this by creating new test questions on the fly that keep the important parts of the question the same, but change the rest to see how well the LLM really understands what it’s being asked. |
