Mapping Overlaps in Benchmarks through Perplexity in the Wild

• URL: http://arxiv.org/abs/2509.23488v3
• Date: Mon, 03 Nov 2025 02:18:28 GMT
• Title: Mapping Overlaps in Benchmarks through Perplexity in the Wild
• Authors: Siyang Wu, Honglin Bao, Sida Li, Ari Holtzman, James A. Evans,
• Abstract summary: We develop signatures of capacity familiarity to characterize large language model (LLM) benchmarks and their meaningful overlaps.<n>Our analysis situates signatures in relation to both the semantic similarity of benchmark questions and the correlation of model performance.<n>Ultimately, we identify cross-functional overlaps across logic, math, language, instruction following, and world modeling, with coding emerging as the least overlapping domain.
• Score: 8.321258152814986
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We develop signatures of capacity familiarity to characterize large language model (LLM) benchmarks and their meaningful overlaps. Benchmark signatures probe the capacity required for benchmark performance. We formally define them as a set of salient tokens drawn from in-the-wild, naturally authored corpora, where LLM token perplexity, reflecting more or less pre-training exposure, becomes highly predictive of LLM benchmark performance. Through a large-scale meta-evaluation, we extract benchmark signatures via stepwise forward selection with linear regressions across 32 LLMs and 88 benchmarks spanning diverse knowledge, coding, logic, instruction following, math, language, reasoning, and world modeling. Our analysis situates signatures in relation to both the semantic similarity of benchmark questions and the correlation of model performance. While performance overlaps are universally high and semantic overlaps remain confined to a narrow mid-range, benchmark signatures prove highly informative in capturing variation, overlap, and divergence. We observe overlap in knowledge and reasoning subtasks, whereas multilingual and cultural benchmarks exhibit less similarity, even compared to cross-task overlap. Notably, performance-level results are strongly influenced by benchmark-orthogonal factors such as question format, highlighting limitations in LLM generalization, the conflation of performance with ability, and issues inherent in current mainstream benchmark agreement studies. Benchmark signatures, however, remain robust to such effects. Ultimately, we identify cross-functional overlaps across logic, math, language, instruction following, and world modeling, with coding emerging as the least overlapping domain. Together, these findings provide mechanistic insights into benchmark validity and LLM sensitivities, and sketch the underlying landscape of interconnected LLM capabilities.

Related papers

• PACIFIC: a framework for generating benchmarks to check Precise Automatically Checked Instruction Following In Code [1.1164117387254457]
  Large Language Model (LLM)-based code assistants have emerged as a powerful application of generative AI.<n>Key requirement for these systems is their ability to accurately follow user instructions.<n>We present PACIFIC, a novel framework designed to automatically generate benchmarks that rigorously assess sequential instruction-following and code dry-running capabilities.
  arXiv  Detail & Related papers  (2025-12-11T14:49:56Z)
• Beyond Synthetic Benchmarks: Evaluating LLM Performance on Real-World Class-Level Code Generation [3.9189409002585567]
  Large language models (LLMs) have demonstrated strong performance on function-level code generation benchmarks.<n>We introduce a benchmark derived from real-world open-source repositories to evaluate generalization under practical conditions.<n>We examine how input specification completeness and retrieval-augmented generation affect class-level correctness across multiple state-of-the-art LLMs.
  arXiv  Detail & Related papers  (2025-10-30T04:30:23Z)
• DAG-Math: Graph-Guided Mathematical Reasoning in LLMs [54.231935013127206]
  Large Language Models (LLMs) demonstrate strong performance on mathematical problems when prompted with Chain-of-Thought (CoT)<n>We propose modeling CoT as a certain rule-based process over directed acyclic graphs (DAGs)<n>We introduce logical closeness, a metric that quantifies how well a model's CoT trajectory adheres to the DAG structure.
  arXiv  Detail & Related papers  (2025-10-19T21:05:17Z)
• Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models [82.87985794856803]
  Large Language Models (LLMs) have achieved state-of-the-art performance on a broad range of Natural Language Processing (NLP) tasks.<n>Recently, Diffusion Language Models (DLMs) have emerged as a promising alternative architecture.
  arXiv  Detail & Related papers  (2025-10-05T10:50:52Z)
• Re-Evaluating Code LLM Benchmarks Under Semantic Mutation [8.58692613099365]
  We present an empirical study to investigate prompt sensitivity in code benchmarks.<n>We propose a general framework that modifies prompt templates in a manner that preserves both their semantics and their structure.<n>Our findings suggest that even slight prompt variations can lead to significant shifts in performance.
  arXiv  Detail & Related papers  (2025-06-20T15:30:36Z)
• Aligning MLLM Benchmark With Human Preferences via Structural Equation Modeling [17.092510377905814]
  evaluating multimodal large language models (MLLMs) remains a fundamental challenge due to a lack of structured, interpretable, and theoretically grounded benchmark designs.<n>We propose a novel framework for aligning MLLM benchmark based on Structural Equation Modeling (SEM) to analyze and quantify the internal validity, dimensional separability, and contribution of benchmark components.<n> Experimental results demonstrate that the proposed benchmark exhibits stronger interpretability, reduced indicator redundancy, and clearer cognitive consistency compared to existing approaches.
  arXiv  Detail & Related papers  (2025-06-13T08:04:56Z)
• IDA-Bench: Evaluating LLMs on Interactive Guided Data Analysis [60.32962597618861]
  IDA-Bench is a novel benchmark evaluating large language models in multi-round interactive scenarios.<n>Agent performance is judged by comparing its final numerical output to the human-derived baseline.<n>Even state-of-the-art coding agents (like Claude-3.7-thinking) succeed on 50% of the tasks, highlighting limitations not evident in single-turn tests.
  arXiv  Detail & Related papers  (2025-05-23T09:37:52Z)
• Model Utility Law: Evaluating LLMs beyond Performance through Mechanism Interpretable Metric [99.56567010306807]
  Large Language Models (LLMs) have become indispensable across academia, industry, and daily applications.<n>One core challenge of evaluation in the large language model (LLM) era is the generalization issue.<n>We propose Model Utilization Index (MUI), a mechanism interpretability enhanced metric that complements traditional performance scores.
  arXiv  Detail & Related papers  (2025-04-10T04:09:47Z)
• Beyond the Singular: The Essential Role of Multiple Generations in Effective Benchmark Evaluation and Analysis [10.133537818749291]
  Large language models (LLMs) have demonstrated significant utilities in real-world applications.<n> Benchmark evaluations are crucial for assessing the capabilities of LLMs.
  arXiv  Detail & Related papers  (2025-02-13T03:43:33Z)
• DARG: Dynamic Evaluation of Large Language Models via Adaptive Reasoning Graph [70.79413606968814]
  We introduce Dynamic Evaluation of LLMs via Adaptive Reasoning Graph Evolvement (DARG) to dynamically extend current benchmarks with controlled complexity and diversity. Specifically, we first extract the reasoning graphs of data points in current benchmarks and then perturb the reasoning graphs to generate novel testing data. Such newly generated test samples can have different levels of complexity while maintaining linguistic diversity similar to the original benchmarks.
  arXiv  Detail & Related papers  (2024-06-25T04:27:53Z)
• MR-Ben: A Meta-Reasoning Benchmark for Evaluating System-2 Thinking in LLMs [55.20845457594977]
  Large language models (LLMs) have shown increasing capability in problem-solving and decision-making.<n>We present a process-based benchmark MR-Ben that demands a meta-reasoning skill.<n>Our meta-reasoning paradigm is especially suited for system-2 slow thinking.
  arXiv  Detail & Related papers  (2024-06-20T03:50:23Z)
• NPHardEval4V: Dynamic Evaluation of Large Vision-Language Models with Effects of Vision [64.83085920775316]
  We introduce NPHardEval4V, a multimodal benchmark suite grounded in four classical NP-hard problems.<n>Each task is presented through a combination of structured visual layouts and textual prompts, designed to assess the ability of LVLMs to perform reasoning under visual-linguistic constraints.<n>Our results show that while these models perform reasonably well on perception-based inputs, they struggle with global optimization, abstraction, and constraint satisfaction.
  arXiv  Detail & Related papers  (2024-03-04T07:10:31Z)
• The Benchmark Lottery [114.43978017484893]
  "A benchmark lottery" describes the overall fragility of the machine learning benchmarking process. We show that the relative performance of algorithms may be altered significantly simply by choosing different benchmark tasks.
  arXiv  Detail & Related papers  (2021-07-14T21:08:30Z)

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.