Why Global LLM Leaderboards Are Misleading: Small Portfolios for Heterogeneous Supervised ML

• URL: http://arxiv.org/abs/2605.06656v1
• Date: Thu, 07 May 2026 17:57:58 GMT
• Title: Why Global LLM Leaderboards Are Misleading: Small Portfolios for Heterogeneous Supervised ML
• Abstract summary: We analyze 89K comparisons in 116 languages from 52 LLMs from Arena.<n>Nearly 2/3 of the decisive votes cancel out, and even the top 50 models according to the global Bradley-Terry ranking are statistically indistinguishable.<n>We introduce the framework of $(, )$-portfolios, which are small sets of models that achieve a prediction error at most $$, "covering" at least a $$ fraction of users.
• Score: 1.373282478189168
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Ranking LLMs via pairwise human feedback underpins current leaderboards for open-ended tasks, such as creative writing and problem-solving. We analyze ~89K comparisons in 116 languages from 52 LLMs from Arena, and show that the best-fit global Bradley-Terry (BT) ranking is misleading. Nearly 2/3 of the decisive votes cancel out, and even the top 50 models according to the global BT ranking are statistically indistinguishable (pairwise win probabilities are at most 0.53 within the top 50 models). We trace this failure to strong, structured heterogeneity of opinions across language, task, and time. Moreover, we find an important characteristic - *language* plays a key role. Grouping by language (and families) increases the agreement of votes massively, resulting in two orders of magnitude higher spread in the ELO scores (i.e., very consistent rankings). What appears as global noise is in fact a mixture of coherent but conflicting subpopulations. To address such heterogeneity in supervised machine learning, we introduce the framework of $(λ, ν)$-portfolios, which are small sets of models that achieve a prediction error at most $λ$, "covering" at least a $ν$ fraction of users. We formulate this as a variant of the set cover problem and provide guarantees using the VC dimension of the underlying set system. On the Arena data, our algorithms recover just 5 distinct BT rankings that cover over 96% of votes at a modest $λ$, compared to the 21% coverage by the global ranking. We also provide a portfolio of 6 LLMs that cover twice as many votes as the top-6 LLMs from a global ranking. We further construct portfolios for a classification problem on the COMPAS dataset using an ensemble of fairness-regularized classification models and show that these portfolios can be used to detect blind spots in the data, which might be of independent interest to policymakers.

Related papers

• Whose Gold? Annotator-Pool Disagreement Is Large at the Item Level, and Hidden by Small Leaderboards [0.0]
  We show that preference benchmarks are built by hiring annotators, and the identity of those annotators is treated as an implementation detail.<n>On the 2,885 MultiPref items where both pools are internally unanimous, so no tie-breaking convention is consulted at all.<n>Yet on both corpora the resulting model leaderboards are bit-identical: Kendall tau = 1.00 with zero of six models displaced.
  arXiv  Detail & Related papers  (2026-08-17T00:19:52Z)
• Knowledge Index of Noah's Ark [63.143852586221534]
  We introduce KINA, an 899-item benchmark across 261 fine-grained disciplines.<n>We prove a bonus-on-bar tournament weakly FOSD-dominates flat payment in released-review quality.<n>The top model, Gemini-3.1-Pro-Preview, reaches 53.17%, followed by Claude-Opus-4.6 at 49.92% and GPT-5.4 at 48.55%.
  arXiv  Detail & Related papers  (2026-06-03T17:06:49Z)
• LRanker: LLM Ranker for Massive Candidates [32.698478901429375]
  We propose LRanker, a framework tailored for large-candidate ranking.<n>LRanker incorporates a candidate aggregation encoder that leverages K-means clustering to explicitly model global candidate information.<n>We evaluate LRanker on seven tasks across three scenarios in RBench with different candidate scales.
  arXiv  Detail & Related papers  (2026-05-27T01:04:36Z)
• Do We Still Need Humans in the Loop? Comparing Human and LLM Annotation in Active Learning for Hostility Detection [68.37351671559675]
  Active learning can annotate thousands of instances from a short prompt at negligible cost.<n>This raises two questions for active learning (AL): can LLM labels replace human labels within the AL loop, and does AL remain necessary when entire corpora can be labelled at once?<n>We investigate both questions on a new dataset of 277,902 German political TikTok comments.
  arXiv  Detail & Related papers  (2026-04-15T14:10:58Z)
• Can Persona-Prompted LLMs Emulate Subgroup Values? An Empirical Analysis of Generalisability and Fairness in Cultural Alignment [75.88231994139132]
  We show that even state-of-the-art models like GPT-4.1 achieve only 57.4% accuracy in predicting modal preferences.<n>We find significant pre-existing performance biases, where models better emulate young, male, Chinese, and Christian personas.
  arXiv  Detail & Related papers  (2026-04-14T15:06:13Z)
• BLITZRANK: Principled Zero-shot Ranking Agents with Tournament Graphs [14.085089126904101]
  We introduce a tournament graph framework that provides a principled foundation for $k$-wise ranking.<n>Each $k$-item comparison reveals a complete tournament of $binomk2$ pairwise preferences.<n>We formalize when an item's rank is certifiably determined and design a greedy query schedule that maximizes information gain.
  arXiv  Detail & Related papers  (2026-02-05T08:41:00Z)
• Evaluating the Promise and Pitfalls of LLMs in Hiring Decisions [1.1883838320818292]
  Large language models (LLMs) in hiring promise to streamline candidate screening, but it also raises serious concerns regarding accuracy and algorithmic bias.<n>We benchmark several state-of-the-art foundational LLMs and compare them with our proprietary domain-specific hiring model (Match Score) for job candidate matching.<n>Our experiments show that Match Score outperforms the general-purpose LLMs on accuracy (ROC AUC 0.85 vs 0.77) and achieves significantly more equitable outcomes across demographic groups.
  arXiv  Detail & Related papers  (2025-07-02T19:02:18Z)
• GLIDER: Grading LLM Interactions and Decisions using Explainable Ranking [0.9614204956530676]
  We introduce GLIDER, a powerful 3B evaluator LLM that can score any text input and associated context on arbitrary user defined criteria.<n>GLIDER shows higher Pearson's correlation than GPT-4o on FLASK and greatly outperforms prior evaluation models.<n>It supports fine-grained scoring, multilingual reasoning, span highlighting and was trained on 685 domains and 183 criteria.
  arXiv  Detail & Related papers  (2024-12-18T18:41:12Z)
• Confidence Diagram of Nonparametric Ranking for Uncertainty Assessment in Large Language Models Evaluation [20.022623972491733]
  Ranking large language models (LLMs) has proven to be an effective tool to improve alignment based on the best-of-$N$ policy.<n>We propose a new inferential framework for hypothesis testing among the ranking for language models.
  arXiv  Detail & Related papers  (2024-12-07T02:34:30Z)
• Toward Automatic Relevance Judgment using Vision--Language Models for Image--Text Retrieval Evaluation [56.49084589053732]
  Vision--Language Models (VLMs) have demonstrated success across diverse applications, yet their potential to assist in relevance judgments remains uncertain. This paper assesses the relevance estimation capabilities of VLMs, including CLIP, LLaVA, and GPT-4V, within a large-scale textitad hoc retrieval task tailored for multimedia content creation in a zero-shot fashion.
  arXiv  Detail & Related papers  (2024-08-02T16:15:25Z)
• Paraphrase and Aggregate with Large Language Models for Minimizing Intent Classification Errors [19.601600598570215]
  We show that large language models (LLM) can achieve high performance on large multi-class classification tasks but still make classification errors and worse, generate out-of-vocabulary class labels. We introduce Paraphrase and AGgregate (PAG)-LLM approach wherein an LLM generates multiple paraphrases of the input query (parallel queries) We show that PAG-LLM is especially effective for hard examples where LLM is uncertain, and reduces the critical misclassification and hallucinated label generation errors.
  arXiv  Detail & Related papers  (2024-06-24T22:30:26Z)
• "Knowing When You Don't Know": A Multilingual Relevance Assessment Dataset for Robust Retrieval-Augmented Generation [90.09260023184932]
  Retrieval-Augmented Generation (RAG) grounds Large Language Model (LLM) output by leveraging external knowledge sources to reduce factual hallucinations. NoMIRACL is a human-annotated dataset for evaluating LLM robustness in RAG across 18 typologically diverse languages. We measure relevance assessment using: (i) hallucination rate, measuring model tendency to hallucinate, when the answer is not present in passages in the non-relevant subset, and (ii) error rate, measuring model inaccuracy to recognize relevant passages in the relevant subset.
  arXiv  Detail & Related papers  (2023-12-18T17:18:04Z)
• Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting [65.00288634420812]
  Pairwise Ranking Prompting (PRP) is a technique to significantly reduce the burden on Large Language Models (LLMs) Our results are the first in the literature to achieve state-of-the-art ranking performance on standard benchmarks using moderate-sized open-sourced LLMs.
  arXiv  Detail & Related papers  (2023-06-30T11:32:25Z)
• Is ChatGPT Good at Search? Investigating Large Language Models as Re-Ranking Agents [53.78782375511531]
  Large Language Models (LLMs) have demonstrated remarkable zero-shot generalization across various language-related tasks.<n>This paper investigates generative LLMs for relevance ranking in Information Retrieval (IR)<n>To address concerns about data contamination of LLMs, we collect a new test set called NovelEval.<n>To improve efficiency in real-world applications, we delve into the potential for distilling the ranking capabilities of ChatGPT into small specialized models.
  arXiv  Detail & Related papers  (2023-04-19T10:16:03Z)

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.