Information alignment evaluators are vital for various NLG evaluation tasks and trustworthy LLM deployment, reducing hallucinations and enhancing user trust. Current fine-grained methods, like FactScore, verify facts individually but neglect inter-fact dependencies, enabling subtle vulnerabilities. In this work, we introduce MontageLie, a challenging benchmark that constructs deceptive narratives by "montaging" truthful statements without introducing explicit hallucinations. We demonstrate that both coarse-grained LLM-based evaluators and current fine-grained frameworks are susceptible to this attack, with AUC-ROC scores falling below 65%. To enable more robust fine-grained evaluation, we propose DoveScore, a novel framework that jointly verifies factual accuracy and event-order consistency. By modeling inter-fact relationships, DoveScore outperforms existing fine-grained methods by over 8%, providing a more robust solution for long-form text alignment evaluation. Our code and datasets are available at https://github.com/dannalily/DoveScore.
Evaluating natural language generation (NLG) systems is challenging due to the diversity of valid outputs. While human evaluation is the gold standard, it suffers from inconsistencies, lack of standardisation, and demographic biases, limiting reproducibility. LLM-based evaluation offers a scalable alternative but is highly sensitive to prompt design, where small variations can lead to significant discrepancies. In this work, we propose an inversion learning method that learns effective reverse mappings from model outputs back to their input instructions, enabling the automatic generation of highly effective, model-specific evaluation prompts. Our method requires only a single evaluation sample and eliminates the need for time-consuming manual prompt engineering, thereby improving both efficiency and robustness. Our work contributes toward a new direction for more robust and efficient LLM-based evaluation.
Reasoning-enabled large language models (LLMs) have recently demonstrated impressive performance in complex logical and mathematical tasks, yet their effectiveness in evaluating natural language generation remains unexplored. This study systematically compares reasoning-based LLMs (DeepSeek-R1 and OpenAI o3) with their non-reasoning counterparts across machine translation (MT) and text summarization (TS) evaluation tasks. We evaluate eight models across three architectural categories, including state-of-the-art reasoning models, their distilled variants (ranging from 8B to 70B parameters), and equivalent conventional, non-reasoning LLMs. Our experiments on WMT23 and SummEval benchmarks reveal that the benefits of reasoning capabilities are highly model and task-dependent: while OpenAI o3-mini models show consistent performance improvements with increased reasoning intensity, DeepSeek-R1 underperforms compared to its non-reasoning variant, with exception to certain aspects of TS evaluation. Correlation analysis demonstrates that increased reasoning token usage positively correlates with evaluation quality in o3-mini models. Furthermore, our results show that distillation of reasoning capabilities maintains reasonable performance in medium-sized models (32B) but degrades substantially in smaller variants (8B). This work provides the first comprehensive assessment of reasoning LLMs for NLG evaluation and offers insights into their practical use.
Previous research has shown that LLMs have potential in multilingual NLG evaluation tasks. However, existing research has not fully explored the differences in the evaluation capabilities of LLMs across different languages. To this end, this study provides a comprehensive analysis of the multilingual evaluation performance of 10 recent LLMs, spanning high-resource and low-resource languages through correlation analysis, perturbation attacks, and fine-tuning. We found that 1) excluding the reference answer from the prompt and using large-parameter LLM-based evaluators leads to better performance across various languages; 2) most LLM-based evaluators show a higher correlation with human judgments in high-resource languages than in low-resource languages; 3) in the languages where they are most sensitive to such attacks, they also tend to exhibit the highest correlation with human judgments; and 4) fine-tuning with data from a particular language yields a broadly consistent enhancement in the model's evaluation performance across diverse languages. Our findings highlight the imbalance in LLMs'evaluation capabilities across different languages and suggest that low-resource language scenarios deserve more attention.
Large Language Models (LLMs) have demonstrated great potential as evaluators of NLG systems, allowing for high-quality, reference-free, and multi-aspect assessments. However, existing LLM-based metrics suffer from two major drawbacks: reliance on proprietary models to generate training data or perform evaluations, and a lack of fine-grained, explanatory feedback. In this paper, we introduce OpeNLGauge, a fully open-source, reference-free NLG evaluation metric that provides accurate explanations based on error spans. OpeNLGauge is available as a two-stage ensemble of larger open-weight LLMs, or as a small fine-tuned evaluation model, with confirmed generalizability to unseen tasks, domains and aspects. Our extensive meta-evaluation shows that OpeNLGauge achieves competitive correlation with human judgments, outperforming state-of-the-art models on certain tasks while maintaining full reproducibility and providing explanations more than twice as accurate.
Large Language Model (LLM) integrations into applications like Microsoft365 suite and Google Workspace for creating/processing documents, emails, presentations, etc. has led to considerable enhancements in productivity and time savings. But as these integrations become more more complex, it is paramount to ensure that the quality of output from the LLM-integrated applications are relevant and appropriate for use. Identifying the need to develop robust evaluation approaches for natural language generation, wherein references/ground labels doesn't exist or isn't amply available, this paper introduces a novel framework called "SAGEval" which utilizes a critiquing Agent to provide feedback on scores generated by LLM evaluators. We show that the critiquing Agent is able to rectify scores from LLM evaluators, in absence of references/ground-truth labels, thereby reducing the need for labeled data even for complex NLG evaluation scenarios, like the generation of JSON-structured forms/surveys with responses in different styles like multiple choice, likert ratings, single choice questions, etc.
Current automated fact-checking (AFC) approaches commonly evaluate evidence either implicitly via the predicted verdicts or by comparing retrieved evidence with a predefined closed knowledge source, such as Wikipedia. However, these methods suffer from limitations, resulting from their reliance on evaluation metrics developed for different purposes and constraints imposed by closed knowledge sources. Recent advances in natural language generation (NLG) evaluation offer new possibilities for evidence assessment. In this work, we introduce Ev2R, an evaluation framework for AFC that comprises three types of approaches for evidence evaluation: reference-based, proxy-reference, and reference-less. We evaluate their effectiveness through agreement with human ratings and adversarial tests, and demonstrate that prompt-based scorers, particularly those leveraging LLMs and reference evidence, outperform traditional evaluation approaches.
The correlation between NLG automatic evaluation metrics and human evaluation is often regarded as a critical criterion for assessing the capability of an evaluation metric. However, different grouping methods and correlation coefficients result in various types of correlation measures used in meta-evaluation. In specific evaluation scenarios, prior work often directly follows conventional measure settings, but the characteristics and differences between these measures have not gotten sufficient attention. Therefore, this paper analyzes 12 common correlation measures using a large amount of real-world data from six widely-used NLG evaluation datasets and 32 evaluation metrics, revealing that different measures indeed impact the meta-evaluation results. Furthermore, we propose three perspectives that reflect the capability of meta-evaluation: discriminative power, ranking consistency, and sensitivity to score granularity. We find that the measure using global grouping and Pearson correlation coefficient exhibits the best performance in both discriminative power and ranking consistency. Besides, the measures using system-level grouping or Kendall correlation are the least sensitive to score granularity.
The conventional paradigm of using large language models (LLMs) for evaluating natural language generation (NLG) systems typically relies on two key inputs: (1) a clear definition of the NLG task to be evaluated and (2) a list of pre-defined evaluation criteria. This process treats LLMs as ''passive critics,'' strictly following human-defined criteria for evaluation. However, as new NLG tasks emerge, the criteria for assessing text quality can vary greatly. Consequently, these rigid evaluation methods struggle to adapt to diverse NLG tasks without extensive prompt engineering customized for each specific task. To address this limitation, we introduce Active-Critic, a novel LLM-based NLG evaluation protocol that enables LLMs to function as ''active critics.'' Specifically, our protocol comprises two key stages. In the first stage, the LLM is instructed to infer the target NLG task and establish relevant evaluation criteria from the data. Building on this self-inferred information, the second stage dynamically optimizes the prompt to guide the LLM toward more human-aligned scoring decisions, while also generating detailed explanations to justify its evaluations. Experiments across four NLG evaluation tasks show that our approach achieves stronger alignment with human judgments than state-of-the-art evaluation methods. Our comprehensive analysis further highlights the effectiveness and explainability of Active-Critic with only a small amount of labeled data. We will share our code and data on GitHub.
Large Language Models (LLMs) are increasingly serving as evaluators in Natural Language Generation (NLG) tasks. However, the capabilities of LLMs in scoring NLG quality remain inadequately explored. Current studies depend on human assessments and simple metrics that fail to capture the discernment of LLMs across diverse NLG tasks. To address this gap, we propose the Discernment of Hierarchical Perturbation (DHP) benchmarking framework, which provides quantitative discernment scores for LLMs utilizing hierarchically perturbed text data and statistical tests to measure the NLG evaluation capabilities of LLMs systematically. We have re-established six evaluation datasets for this benchmark, covering four NLG tasks: Summarization, Story Completion, Question Answering, and Translation. Our comprehensive benchmarking of five major LLM series provides critical insight into their strengths and limitations as NLG evaluators.
Assessing the quality of outputs generated by generative models, such as large language models and vision language models, presents notable challenges. Traditional methods for evaluation typically rely on either human assessments, which are resource-intensive, or automatic metrics that often show a low correlation with human judgment. Another common approach is to use deep learning systems, which not only consume a substantial amount of compute and time but also require extensive training data. In this study, we introduce a tuning-free framework called ReFeR, designed to evaluate generative outputs, including both text and images, by leveraging a 2-level hierarchy of LLMs and VLMs themselves. We rigorously evaluate our framework, ReFeR, across four diverse evaluation tasks. The framework not only improves the accuracy of these evaluations, surpassing previous benchmarks but also generates constructive feedback. Interestingly, the framework is also applicable to reasoning tasks. Experiments on four reasoning tasks demonstrate superior collective reasoning abilities of the framework. We present two variants of the framework: ReFeR-Turbo, optimized for accelerated performance, and ReFeR-Lite, offering a more cost-effective solution. ReFeR-Lite is $\sim7.7\times$ more efficient while being comparably accurate to ReFeR-Turbo. We make code, data and PIP package publicly available. See this PIP URL https://pypi.org/project/refer-agents/ and this Git URL https://github.com/yaswanth-iitkgp/ReFeR_Code .
The evaluation of natural language generation (NLG) tasks is a significant and longstanding research area. With the recent emergence of powerful large language models (LLMs), some studies have turned to LLM-based automatic evaluation methods, which demonstrate great potential to become a new evaluation paradigm following traditional string-based and model-based metrics. However, despite the improved performance of existing methods, they still possess some deficiencies, such as dependency on references and limited evaluation flexibility. Therefore, in this paper, we meticulously construct a large-scale NLG evaluation corpus NLG-Eval with annotations from both human and GPT-4 to alleviate the lack of relevant data in this field. Furthermore, we propose Themis, an LLM dedicated to NLG evaluation, which has been trained with our designed multi-perspective consistency verification and rating-oriented preference alignment methods. Themis can conduct flexible and interpretable evaluations without references, and it exhibits superior evaluation performance on various NLG tasks, simultaneously generalizing well to unseen tasks and surpassing other evaluation models, including GPT-4.
Human evaluation is viewed as a reliable evaluation method for NLG which is expensive and time-consuming. To save labor and costs, researchers usually perform human evaluation on a small subset of data sampled from the whole dataset in practice. However, different selection subsets will lead to different rankings of the systems. To give a more correct inter-system ranking and make the gold standard human evaluation more reliable, we propose a Constrained Active Sampling Framework (CASF) for reliable human judgment. CASF operates through a Learner, a Systematic Sampler and a Constrained Controller to select representative samples for getting a more correct inter-system ranking.Experiment results on 137 real NLG evaluation setups with 44 human evaluation metrics across 16 datasets and 5 NLG tasks demonstrate CASF receives 93.18% top-ranked system recognition accuracy and ranks first or ranks second on 90.91% of the human metrics with 0.83 overall inter-system ranking Kendall correlation.Code and data are publicly available online.
Human evaluation serves as the gold standard for assessing the quality of Natural Language Generation (NLG) systems. Nevertheless, the evaluation guideline, as a pivotal element ensuring reliable and reproducible human assessment, has received limited attention.Our investigation revealed that only 29.84% of recent papers involving human evaluation at top conferences release their evaluation guidelines, with vulnerabilities identified in 77.09% of these guidelines. Unreliable evaluation guidelines can yield inaccurate assessment outcomes, potentially impeding the advancement of NLG in the right direction. To address these challenges, we take an initial step towards reliable evaluation guidelines and propose the first human evaluation guideline dataset by collecting annotations of guidelines extracted from existing papers as well as generated via Large Language Models (LLMs). We then introduce a taxonomy of eight vulnerabilities and formulate a principle for composing evaluation guidelines. Furthermore, a method for detecting guideline vulnerabilities has been explored using LLMs, and we offer a set of recommendations to enhance reliability in human evaluation. The annotated human evaluation guideline dataset and code for the vulnerability detection method are publicly available online.
The automatic evaluation of natural language generation (NLG) systems presents a long-lasting challenge. Recent studies have highlighted various neural metrics that align well with human evaluations. Yet, the robustness of these evaluators against adversarial perturbations remains largely under-explored due to the unique challenges in obtaining adversarial data for different NLG evaluation tasks. To address the problem, we introduce AdvEval, a novel black-box adversarial framework against NLG evaluators. AdvEval is specially tailored to generate data that yield strong disagreements between human and victim evaluators. Specifically, inspired by the recent success of large language models (LLMs) in text generation and evaluation, we adopt strong LLMs as both the data generator and gold evaluator. Adversarial data are automatically optimized with feedback from the gold and victim evaluator. We conduct experiments on 12 victim evaluators and 11 NLG datasets, spanning tasks including dialogue, summarization, and question evaluation. The results show that AdvEval can lead to significant performance degradation of various victim metrics, thereby validating its efficacy.
As natural language generation (NLG) models have become prevalent, systematically assessing the quality of machine-generated texts has become increasingly important. Recent studies introduce LLM-based evaluators that operate as reference-free metrics, demonstrating their capability to adeptly handle novel tasks. However, these models generally rely on a single-agent approach, which, we argue, introduces an inherent limit to their performance. This is because there exist biases in LLM agent's responses, including preferences for certain text structure or content. In this work, we propose DEBATE, an NLG evaluation framework based on multi-agent scoring system augmented with a concept of Devil's Advocate. Within the framework, one agent is instructed to criticize other agents' arguments, potentially resolving the bias in LLM agent's answers. DEBATE substantially outperforms the previous state-of-the-art methods in two meta-evaluation benchmarks in NLG evaluation, SummEval and TopicalChat. We also show that the extensiveness of debates among agents and the persona of an agent can influence the performance of evaluators.
Watermarking generative-AI systems, such as LLMs, has gained considerable interest, driven by their enhanced capabilities across a wide range of tasks. Although current approaches have demonstrated that small, context-dependent shifts in the word distributions can be used to apply and detect watermarks, there has been little work in analyzing the impact that these perturbations have on the quality of generated texts. Balancing high detectability with minimal performance degradation is crucial in terms of selecting the appropriate watermarking setting; therefore this paper proposes a simple analysis framework where comparative assessment, a flexible NLG evaluation framework, is used to assess the quality degradation caused by a particular watermark setting. We demonstrate that our framework provides easy visualization of the quality-detection trade-off of watermark settings, enabling a simple solution to find an LLM watermark operating point that provides a well-balanced performance. This approach is applied to two different summarization systems and a translation system, enabling cross-model analysis for a task, and cross-task analysis.
Evaluation of opinion summaries using conventional reference-based metrics rarely provides a holistic evaluation and has been shown to have a relatively low correlation with human judgments. Recent studies suggest using Large Language Models (LLMs) as reference-free metrics for NLG evaluation, however, they remain unexplored for opinion summary evaluation. Moreover, limited opinion summary evaluation datasets inhibit progress. To address this, we release the SUMMEVAL-OP dataset covering 7 dimensions related to the evaluation of opinion summaries: fluency, coherence, relevance, faithfulness, aspect coverage, sentiment consistency, and specificity. We investigate Op-I-Prompt a dimension-independent prompt, and Op-Prompts, a dimension-dependent set of prompts for opinion summary evaluation. Experiments indicate that Op-I-Prompt emerges as a good alternative for evaluating opinion summaries achieving an average Spearman correlation of 0.70 with humans, outperforming all previous approaches. To the best of our knowledge, we are the first to investigate LLMs as evaluators on both closed-source and open-source models in the opinion summarization domain.
Some prior work has shown that LLMs perform well in NLG evaluation for different tasks. However, we discover that LLMs seem to confuse different evaluation criteria, which reduces their reliability. For further verification, we first consider avoiding issues of inconsistent conceptualization and vague expression in existing NLG quality criteria themselves. So we summarize a clear hierarchical classification system for 11 common aspects with corresponding different criteria from previous studies involved. Inspired by behavioral testing, we elaborately design 18 types of aspect-targeted perturbation attacks for fine-grained analysis of the evaluation behaviors of different LLMs. We also conduct human annotations beyond the guidance of the classification system to validate the impact of the perturbations. Our experimental results reveal confusion issues inherent in LLMs, as well as other noteworthy phenomena, and necessitate further research and improvements for LLM-based evaluation.
Evaluating natural language generation (NLG) is a vital but challenging problem in natural language processing. Traditional evaluation metrics mainly capturing content (e.g. n-gram) overlap between system outputs and references are far from satisfactory, and large language models (LLMs) such as ChatGPT have demonstrated great potential in NLG evaluation in recent years. Various automatic evaluation methods based on LLMs have been proposed, including metrics derived from LLMs, prompting LLMs, fine-tuning LLMs, and human-LLM collaborative evaluation. In this survey, we first give a taxonomy of LLM-based NLG evaluation methods, and discuss their pros and cons, respectively. Lastly, we discuss several open problems in this area and point out future research directions.