Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts

• URL: http://arxiv.org/abs/2410.09247v1
• Date: Fri, 11 Oct 2024 20:46:56 GMT
• Title: Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts
• Authors: Jacob Haimes, Cenny Wenner, Kunvar Thaman, Vassil Tashev, Clement Neo, Esben Kran, Jason Schreiber,
• Abstract summary: Training data for many Large Language Models (LLMs) is contaminated with test data. Public benchmark scores do not always accurately assess model properties.
• Score: 0.6282171844772422
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: The training data for many Large Language Models (LLMs) is contaminated with test data. This means that public benchmarks used to assess LLMs are compromised, suggesting a performance gap between benchmark scores and actual capabilities. Ideally, a private holdout set could be used to accurately verify scores. Unfortunately, such datasets do not exist for most benchmarks, and post-hoc construction of sufficiently similar datasets is non-trivial. To address these issues, we introduce a systematic methodology for (i) retrospectively constructing a holdout dataset for a target dataset, (ii) demonstrating the statistical indistinguishability of this retro-holdout dataset, and (iii) comparing LLMs on the two datasets to quantify the performance gap due to the dataset's public availability. Applying these methods to TruthfulQA, we construct and release Retro-Misconceptions, on which we evaluate twenty LLMs and find that some have inflated scores by as much as 16 percentage points. Our results demonstrate that public benchmark scores do not always accurately assess model properties, and underscore the importance of improved data practices in the field.

Related papers

• Training on the Benchmark Is Not All You Need [52.01920740114261]
  We propose a simple and effective data leakage detection method based on the contents of multiple-choice options. Our method is able to work under black-box conditions without access to model training data or weights. We evaluate the degree of data leakage of 31 mainstream open-source LLMs on four benchmark datasets.
  arXiv  Detail & Related papers  (2024-09-03T11:09:44Z)
• Truthful Dataset Valuation by Pointwise Mutual Information [28.63827288801458]
  We propose a new data valuation method that provably guarantees the following: data providers always maximize their expected score by truthfully reporting their observed data. Our method, following the paradigm of proper scoring rules, measures the pointwise mutual information (PMI) of the test dataset and the evaluated dataset.
  arXiv  Detail & Related papers  (2024-05-28T15:04:17Z)
• Benchmarking Benchmark Leakage in Large Language Models [24.015208839742343]
  We introduce a detection pipeline utilizing Perplexity and N-gram accuracy, two simple and scalable metrics that gauge a model's prediction precision on benchmark. We reveal substantial instances of training even test set misuse, resulting in potentially unfair comparisons. We propose the "Benchmark Transparency Card" to encourage clear documentation of benchmark utilization.
  arXiv  Detail & Related papers  (2024-04-29T16:05:36Z)
• Investigating Data Contamination in Modern Benchmarks for Large Language Models [27.479260572913724]
  Recent observations have underscored a disparity between the inflated benchmark scores and the actual performance of LLMs. We study data contamination by proposing two methods tailored for both open-source and proprietary LLMs. We find that certain commercial LLMs could surprisingly guess the missing option in various test sets.
  arXiv  Detail & Related papers  (2023-11-16T11:03:04Z)
• Don't Make Your LLM an Evaluation Benchmark Cheater [142.24553056600627]
  Large language models(LLMs) have greatly advanced the frontiers of artificial intelligence, attaining remarkable improvement in model capacity. To assess the model performance, a typical approach is to construct evaluation benchmarks for measuring the ability level of LLMs. We discuss the potential risk and impact of inappropriately using evaluation benchmarks and misleadingly interpreting the evaluation results.
  arXiv  Detail & Related papers  (2023-11-03T14:59:54Z)
• Data Contamination Through the Lens of Time [21.933771085956426]
  Large language models (LLMs) are often supported by evaluating publicly available benchmarks. This practice raises concerns of data contamination, i.e., evaluating on examples that are explicitly or implicitly included in the training data. We conduct the first thorough longitudinal analysis of data contamination in LLMs by using the natural experiment of training cutoffs in GPT models.
  arXiv  Detail & Related papers  (2023-10-16T17:51:29Z)
• On the Evaluation and Refinement of Vision-Language Instruction Tuning Datasets [71.54954966652286]
  We try to evaluate the Vision-Language Instruction-Tuning (VLIT) datasets. We build a new dataset, REVO-LION, by collecting samples with higher SQ from each dataset. Remarkably, even with only half of the complete data, the model trained on REVO-LION can achieve the performance comparable to simply adding all VLIT datasets up.
  arXiv  Detail & Related papers  (2023-10-10T13:01:38Z)
• LLMs as Factual Reasoners: Insights from Existing Benchmarks and Beyond [135.8013388183257]
  We propose a new protocol for inconsistency detection benchmark creation and implement it in a 10-domain benchmark called SummEdits. Most LLMs struggle on SummEdits, with performance close to random chance. The best-performing model, GPT-4, is still 8% below estimated human performance.
  arXiv  Detail & Related papers  (2023-05-23T21:50:06Z)
• DataPerf: Benchmarks for Data-Centric AI Development [81.03754002516862]
  DataPerf is a community-led benchmark suite for evaluating ML datasets and data-centric algorithms. We provide an open, online platform with multiple rounds of challenges to support this iterative development. The benchmarks, online evaluation platform, and baseline implementations are open source.
  arXiv  Detail & Related papers  (2022-07-20T17:47:54Z)
• GEMv2: Multilingual NLG Benchmarking in a Single Line of Code [161.1761414080574]
  Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.
  arXiv  Detail & Related papers  (2022-06-22T17:52:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.