In Natural Language Processing (NLP), the Elo rating system, originally designed for ranking players in dynamic games such as chess, is increasingly being used to evaluate Large Language Models (LLMs) through "A vs B" paired comparisons.However, while popular, the system's suitability for assessing entities with constant skill levels, such as LLMs, remains relatively unexplored. We study two fundamental axioms that evaluation methods should adhere to: reliability and transitivity. We conduct an extensive evaluation of Elo behavior across simulated and real-world scenarios, demonstrating that individual Elo computations can exhibit significant volatility.We show that both axioms are not always satisfied, raising questions about the reliability of current comparative evaluations of LLMs.If the current use of Elo scores is intended to substitute the costly head-to-head comparison of LLMs, it is crucial to ensure the ranking is as robust as possible.Guided by the axioms, our findings offer concrete guidelines for enhancing the reliability of LLM evaluation methods, suggesting a need for reassessment of existing comparative approaches.

Large language models (LLMs) exhibit probabilistic output characteristics, yet conventional evaluation frameworks rely on deterministic scalar metrics. This study introduces a Bayesian approach for LLM capability assessment that integrates prior knowledge through probabilistic inference, addressing limitations under limited-sample regimes. By treating model capabilities as latent variables and leveraging a curated query set to induce discriminative responses, we formalize model ranking as a Bayesian hypothesis testing problem over mutually exclusive capability intervals. Experimental evaluations with GPT-series models demonstrate that the proposed method achieves superior discrimination compared to conventional evaluation methods. Results indicate that even with reduced sample sizes, the approach maintains statistical robustness while providing actionable insights, such as probabilistic statements about a model's likelihood of surpassing specific baselines. This work advances LLM evaluation methodologies by bridging Bayesian inference with practical constraints in real-world deployment scenarios.

With the development of Large Language Models (LLMs), numerous benchmarks have been proposed to measure and compare the capabilities of different LLMs. However, evaluating LLMs is costly due to the large number of test instances and their slow inference speed. In this paper, we aim to explore how to efficiently estimate a model's real performance on a given benchmark based on its evaluation results on a small number of instances sampled from the benchmark. Inspired by Collaborative Filtering (CF) in Recommendation Systems (RS), we treat LLMs as users and test instances as items and propose a two-stage method. In the first stage, we treat instance selection as recommending products to users to choose instances that can easily distinguish model performance. In the second stage, we see performance prediction as rating prediction problem in RS to predict the target LLM's behavior on unselected instances. Experiments on multiple LLMs and datasets imply that our method can accurately estimate the target model's performance while largely reducing its inference overhead.

Recent advancements in Korean large language models (LLMs) have spurred numerous benchmarks and evaluation methodologies, yet the lack of a standardized evaluation framework has led to inconsistent results and limited comparability. To address this, we introduce HRET Haerae Evaluation Toolkit, an open-source, self-evolving evaluation framework tailored specifically for Korean LLMs. HRET unifies diverse evaluation methods, including logit-based scoring, exact-match, language-inconsistency penalization, and LLM-as-a-Judge assessments. Its modular, registry-based architecture integrates major benchmarks (HAE-RAE Bench, KMMLU, KUDGE, HRM8K) and multiple inference backends (vLLM, HuggingFace, OpenAI-compatible endpoints). With automated pipelines for continuous evolution, HRET provides a robust foundation for reproducible, fair, and transparent Korean NLP research.

Generative AI holds significant potential for ecological and environmental applications such as monitoring, data analysis, education, and policy support. However, its effectiveness is limited by the lack of a unified evaluation framework. To address this, we present the Environmental Large Language model Evaluation (ELLE) question answer (QA) dataset, the first benchmark designed to assess large language models and their applications in ecological and environmental sciences. The ELLE dataset includes 1,130 question answer pairs across 16 environmental topics, categorized by domain, difficulty, and type. This comprehensive dataset standardizes performance assessments in these fields, enabling consistent and objective comparisons of generative AI performance. By providing a dedicated evaluation tool, ELLE dataset promotes the development and application of generative AI technologies for sustainable environmental outcomes. The dataset and code are available at https://elle.ceeai.net/ and https://github.com/CEEAI/elle.

Evaluations are critical for understanding the capabilities of large language models (LLMs). Fundamentally, evaluations are experiments; but the literature on evaluations has largely ignored the literature from other sciences on experiment analysis and planning. This article shows researchers with some training in statistics how to think about and analyze data from language model evaluations. Conceptualizing evaluation questions as having been drawn from an unseen super-population, we present formulas for analyzing evaluation data, measuring differences between two models, and planning an evaluation experiment. We make a number of specific recommendations for running language model evaluations and reporting experiment results in a way that minimizes statistical noise and maximizes informativeness.

Despite their sophisticated capabilities, large language models (LLMs) encounter a major hurdle in effective assessment. This paper first revisits the prevalent evaluation method-multiple choice question answering (MCQA), which allows for straightforward accuracy measurement. Through a comprehensive evaluation of 24 models across 11 benchmarks, we highlight several potential drawbacks of MCQA, for instance, the inconsistency between the MCQA evaluation and the generation of open-ended responses in practical scenarios. In response, we introduce an RWQ-Elo rating system, engaging 24 LLMs such as GPT-4, GPT-3.5, Google-Gemini-Pro and LLaMA-1/-2, in a two-player competitive format, with GPT-4 serving as the judge. Each LLM receives an Elo rating thereafter. This system is designed to mirror real-world usage, and for this purpose, we have compiled a new benchmark called ``Real-world questions'' (RWQ), comprising 20,772 authentic user inquiries. Additionally, we thoroughly analyze the characteristics of our system and compare it with prior leaderboards like AlpacaEval and MT-Bench. Our analysis reveals the stability of our RWQ-Elo system, the feasibility of registering new models, and its potential to reshape LLM leaderboards.

Data contamination in evaluation is getting increasingly prevalent with the emergence of language models pre-trained on super large, automatically crawled corpora. This problem leads to significant challenges in the accurate assessment of model capabilities and generalisations. In this paper, we propose LatestEval, an automatic method that leverages the most recent texts to create uncontaminated reading comprehension evaluations. LatestEval avoids data contamination by only using texts published within a recent time window, ensuring no overlap with the training corpora of pre-trained language models. We develop the LatestEval automated pipeline to 1) gather the latest texts; 2) identify key information, and 3) construct questions targeting the information while removing the existing answers from the context. This encourages models to infer the answers themselves based on the remaining context, rather than just copy-paste. Our experiments demonstrate that language models exhibit negligible memorisation behaviours on LatestEval as opposed to previous benchmarks, suggesting a significantly reduced risk of data contamination and leading to a more robust evaluation. Data and code are publicly available at: https://github.com/liyucheng09/LatestEval.

With large language models surpassing human performance on an increasing number of benchmarks, we must take a principled approach for targeted evaluation of model capabilities. Inspired by pseudorandomness, we propose pseudointelligence, which captures the maxim that "(perceived) intelligence lies in the eye of the beholder". That is, that claims of intelligence are meaningful only when their evaluator is taken into account. Concretely, we propose a complexity-theoretic framework of model evaluation cast as a dynamic interaction between a model and a learned evaluator. We demonstrate that this framework can be used to reason about two case studies in language model evaluation, as well as analyze existing evaluation methods.

Data contamination in model evaluation is getting increasingly prevalent as the massive training corpora of large language models often unintentionally include benchmark samples. Therefore, contamination analysis has became an inevitable part of reliable model evaluation. However, existing method of contamination analysis requires the access of the entire training data which is often confidential for recent models. This prevent the community to rigorously audit these models and conduct accurate assessment of their capability. In this paper, we propose a novel method to quantify contamination without the access of the full training set, that measure the extent of contamination with perplexity. Our analysis provides evidence of significant memorisation of recent foundation models in popular reading comprehension, summarisation benchmarks, while multiple choice appears less contaminated.