As Large Language Models (LLMs) are increasingly deployed in sensitive domains such as enterprise and government, ensuring that they adhere to user-defined security policies within context is critical-especially with respect to information non-disclosure. While prior LLM studies have focused on general safety and socially sensitive data, large-scale benchmarks for contextual security preservation against attacks remain lacking. To address this, we introduce a novel large-scale benchmark dataset, CoPriva, evaluating LLM adherence to contextual non-disclosure policies in question answering. Derived from realistic contexts, our dataset includes explicit policies and queries designed as direct and challenging indirect attacks seeking prohibited information. We evaluate 10 LLMs on our benchmark and reveal a significant vulnerability: many models violate user-defined policies and leak sensitive information. This failure is particularly severe against indirect attacks, highlighting a critical gap in current LLM safety alignment for sensitive applications. Our analysis reveals that while models can often identify the correct answer to a query, they struggle to incorporate policy constraints during generation. In contrast, they exhibit a partial ability to revise outputs when explicitly prompted. Our findings underscore the urgent need for more robust methods to guarantee contextual security.
The rapid advancement of large language models (LLMs) has accelerated the development of multi-modal models capable of vocal communication. Unlike text-based interactions, speech conveys rich and diverse information, including semantic content, acoustic variations, paralanguage cues, and environmental context. However, existing evaluations of speech interaction models predominantly focus on the quality of their textual responses, often overlooking critical aspects of vocal performance and lacking benchmarks with vocal-specific test instances. To address this gap, we propose VocalBench, a comprehensive benchmark designed to evaluate speech interaction models' capabilities in vocal communication. VocalBench comprises 9,400 carefully curated instances across four key dimensions: semantic quality, acoustic performance, conversational abilities, and robustness. It covers 16 fundamental skills essential for effective vocal interaction. Experimental results reveal significant variability in current model capabilities, each exhibiting distinct strengths and weaknesses, and provide valuable insights to guide future research in speech-based interaction systems. Code and evaluation instances are available at https://github.com/SJTU-OmniAgent/VocalBench.
Accurately predicting cellular responses to genetic perturbations is essential for understanding disease mechanisms and designing effective therapies. Yet exhaustively exploring the space of possible perturbations (e.g., multi-gene perturbations or across tissues and cell types) is prohibitively expensive, motivating methods that can generalize to unseen conditions. In this work, we explore how knowledge graphs of gene-gene relationships can improve out-of-distribution (OOD) prediction across three challenging settings: unseen single perturbations; unseen double perturbations; and unseen cell lines. In particular, we present: (i) TxPert, a new state-of-the-art method that leverages multiple biological knowledge networks to predict transcriptional responses under OOD scenarios; (ii) an in-depth analysis demonstrating the impact of graphs, model architecture, and data on performance; and (iii) an expanded benchmarking framework that strengthens evaluation standards for perturbation modeling.
Neural Architecture Search (NAS) accelerates progress in deep learning through systematic refinement of model architectures. The downside is increasingly large energy consumption during the search process. Surrogate-based benchmarking mitigates the cost of full training by querying a pre-trained surrogate to obtain an estimate for the quality of the model. Specifically, energy-aware benchmarking aims to make it possible for NAS to favourably trade off model energy consumption against accuracy. Towards this end, we propose three design principles for such energy-aware benchmarks: (i) reliable power measurements, (ii) a wide range of GPU usage, and (iii) holistic cost reporting. We analyse EA-HAS-Bench based on these principles and find that the choice of GPU measurement API has a large impact on the quality of results. Using the Nvidia System Management Interface (SMI) on top of its underlying library influences the sampling rate during the initial data collection, returning faulty low-power estimations. This results in poor correlation with accurate measurements obtained from an external power meter. With this study, we bring to attention several key considerations when performing energy-aware surrogate-based benchmarking and derive first guidelines that can help design novel benchmarks. We show a narrow usage range of the four GPUs attached to our device, ranging from 146 W to 305 W in a single-GPU setting, and narrowing down even further when using all four GPUs. To improve holistic energy reporting, we propose calibration experiments over assumptions made in popular tools, such as Code Carbon, thus achieving reductions in the maximum inaccuracy from 10.3 % to 8.9 % without and to 6.6 % with prior estimation of the expected load on the device.
The rise of IoT has increased the need for on-edge machine learning, with TinyML emerging as a promising solution for resource-constrained devices such as MCU. However, evaluating their performance remains challenging due to diverse architectures and application scenarios. Current solutions have many non-negligible limitations. This work introduces an alternative benchmarking methodology that integrates energy and latency measurements while distinguishing three execution phases pre-inference, inference, and post-inference. Additionally, the setup ensures that the device operates without being powered by an external measurement unit, while automated testing can be leveraged to enhance statistical significance. To evaluate our setup, we tested the STM32N6 MCU, which includes a NPU for executing neural networks. Two configurations were considered: high-performance and Low-power. The variation of the EDP was analyzed separately for each phase, providing insights into the impact of hardware configurations on energy efficiency. Each model was tested 1000 times to ensure statistically relevant results. Our findings demonstrate that reducing the core voltage and clock frequency improve the efficiency of pre- and post-processing without significantly affecting network execution performance. This approach can also be used for cross-platform comparisons to determine the most efficient inference platform and to quantify how pre- and post-processing overhead varies across different hardware implementations.
Leveraging the power of multimodal large language models (LLMs) offers a promising approach to enhancing the accuracy and interpretability of morphing attack detection (MAD), especially in real-world biometric applications. This work introduces the use of LLMs for differential morphing attack detection (D-MAD). To the best of our knowledge, this is the first study to employ multimodal LLMs to D-MAD using real biometric data. To effectively utilize these models, we design Chain-of-Thought (CoT)-based prompts to reduce failure-to-answer rates and enhance the reasoning behind decisions. Our contributions include: (1) the first application of multimodal LLMs for D-MAD using real data subjects, (2) CoT-based prompt engineering to improve response reliability and explainability, (3) comprehensive qualitative and quantitative benchmarking of LLM performance using data from 54 individuals captured in passport enrollment scenarios, and (4) comparative analysis of two multimodal LLMs: ChatGPT-4o and Gemini providing insights into their morphing attack detection accuracy and decision transparency. Experimental results show that ChatGPT-4o outperforms Gemini in detection accuracy, especially against GAN-based morphs, though both models struggle under challenging conditions. While Gemini offers more consistent explanations, ChatGPT-4o is more resilient but prone to a higher failure-to-answer rate.
Malocclusion is a major challenge in orthodontics, and its complex presentation and diverse clinical manifestations make accurate localization and diagnosis particularly important. Currently, one of the major shortcomings facing the field of dental image analysis is the lack of large-scale, accurately labeled datasets dedicated to malocclusion issues, which limits the development of automated diagnostics in the field of dentistry and leads to a lack of diagnostic accuracy and efficiency in clinical practice. Therefore, in this study, we propose the Oral and Maxillofacial Natural Images (OMNI) dataset, a novel and comprehensive dental image dataset aimed at advancing the study of analyzing dental images for issues of malocclusion. Specifically, the dataset contains 4166 multi-view images with 384 participants in data collection and annotated by professional dentists. In addition, we performed a comprehensive validation of the created OMNI dataset, including three CNN-based methods, two Transformer-based methods, and one GNN-based method, and conducted automated diagnostic experiments for malocclusion issues. The experimental results show that the OMNI dataset can facilitate the automated diagnosis research of malocclusion issues and provide a new benchmark for the research in this field. Our OMNI dataset and baseline code are publicly available at https://github.com/RoundFaceJ/OMNI.
Survival analysis, which estimates the probability of event occurrence over time from censored data, is fundamental in numerous real-world applications, particularly in high-stakes domains such as healthcare and risk assessment. Despite advances in numerous survival models, quantifying the uncertainty of predictions from these models remains underexplored and challenging. The lack of reliable uncertainty quantification limits the interpretability and trustworthiness of survival models, hindering their adoption in clinical decision-making and other sensitive applications. To bridge this gap, in this work, we introduce SurvUnc, a novel meta-model based framework for post-hoc uncertainty quantification for survival models. SurvUnc introduces an anchor-based learning strategy that integrates concordance knowledge into meta-model optimization, leveraging pairwise ranking performance to estimate uncertainty effectively. Notably, our framework is model-agnostic, ensuring compatibility with any survival model without requiring modifications to its architecture or access to its internal parameters. Especially, we design a comprehensive evaluation pipeline tailored to this critical yet overlooked problem. Through extensive experiments on four publicly available benchmarking datasets and five representative survival models, we demonstrate the superiority of SurvUnc across multiple evaluation scenarios, including selective prediction, misprediction detection, and out-of-domain detection. Our results highlight the effectiveness of SurvUnc in enhancing model interpretability and reliability, paving the way for more trustworthy survival predictions in real-world applications.
The proliferation of Large Language Models (LLMs) and Intelligent Virtual Agents acting as psychotherapists presents significant opportunities for expanding mental healthcare access. However, their deployment has also been linked to serious adverse outcomes, including user harm and suicide, facilitated by a lack of standardized evaluation methodologies capable of capturing the nuanced risks of therapeutic interaction. Current evaluation techniques lack the sensitivity to detect subtle changes in patient cognition and behavior during therapy sessions that may lead to subsequent decompensation. We introduce a novel risk taxonomy specifically designed for the systematic evaluation of conversational AI psychotherapists. Developed through an iterative process including review of the psychotherapy risk literature, qualitative interviews with clinical and legal experts, and alignment with established clinical criteria (e.g., DSM-5) and existing assessment tools (e.g., NEQ, UE-ATR), the taxonomy aims to provide a structured approach to identifying and assessing user/patient harms. We provide a high-level overview of this taxonomy, detailing its grounding, and discuss potential use cases. We discuss two use cases in detail: monitoring cognitive model-based risk factors during a counseling conversation to detect unsafe deviations, in both human-AI counseling sessions and in automated benchmarking of AI psychotherapists with simulated patients. The proposed taxonomy offers a foundational step towards establishing safer and more responsible innovation in the domain of AI-driven mental health support.
The rapid evolution of multimodal large language models (MLLMs) has significantly enhanced their real-world applications. However, achieving consistent performance across languages, especially when integrating cultural knowledge, remains a significant challenge. To better assess this issue, we introduce two new benchmarks: KnowRecall and VisRecall, which evaluate cross-lingual consistency in MLLMs. KnowRecall is a visual question answering benchmark designed to measure factual knowledge consistency in 15 languages, focusing on cultural and historical questions about global landmarks. VisRecall assesses visual memory consistency by asking models to describe landmark appearances in 9 languages without access to images. Experimental results reveal that state-of-the-art MLLMs, including proprietary ones, still struggle to achieve cross-lingual consistency. This underscores the need for more robust approaches that produce truly multilingual and culturally aware models.
Large reasoning models such as OpenAI o1 and DeepSeek-R1 have achieved remarkable performance in the domain of reasoning. A key component of their training is the incorporation of verifiable rewards within reinforcement learning (RL). However, existing reward benchmarks do not evaluate reference-based reward systems, leaving researchers with limited understanding of the accuracy of verifiers used in RL. In this paper, we introduce two benchmarks, VerifyBench and VerifyBench-Hard, designed to assess the performance of reference-based reward systems. These benchmarks are constructed through meticulous data collection and curation, followed by careful human annotation to ensure high quality. Current models still show considerable room for improvement on both VerifyBench and VerifyBench-Hard, especially smaller-scale models. Furthermore, we conduct a thorough and comprehensive analysis of evaluation results, offering insights for understanding and developing reference-based reward systems. Our proposed benchmarks serve as effective tools for guiding the development of verifier accuracy and the reasoning capabilities of models trained via RL in reasoning tasks.
The evaluation of large language models (LLMs) via benchmarks is widespread, yet inconsistencies between different leaderboards and poor separability among top models raise concerns about their ability to accurately reflect authentic model capabilities. This paper provides a critical analysis of benchmark effectiveness, examining main-stream prominent LLM benchmarks using results from diverse models. We first propose a new framework for accurate and reliable estimations of item characteristics and model abilities. Specifically, we propose Pseudo-Siamese Network for Item Response Theory (PSN-IRT), an enhanced Item Response Theory framework that incorporates a rich set of item parameters within an IRT-grounded architecture. Based on PSN-IRT, we conduct extensive analysis which reveals significant and varied shortcomings in the measurement quality of current benchmarks. Furthermore, we demonstrate that leveraging PSN-IRT is able to construct smaller benchmarks while maintaining stronger alignment with human preference.
Autonomous systems that rely on Machine Learning (ML) utilize online fault tolerance mechanisms, such as runtime monitors, to detect ML prediction errors and maintain safety during operation. However, the lack of human-interpretable explanations for these errors can hinder the creation of strong assurances about the system's safety and reliability. This paper introduces a novel fuzzy-based monitor tailored for ML perception components. It provides human-interpretable explanations about how different operating conditions affect the reliability of perception components and also functions as a runtime safety monitor. We evaluated our proposed monitor using naturalistic driving datasets as part of an automated driving case study. The interpretability of the monitor was evaluated and we identified a set of operating conditions in which the perception component performs reliably. Additionally, we created an assurance case that links unit-level evidence of \textit{correct} ML operation to system-level \textit{safety}. The benchmarking demonstrated that our monitor achieved a better increase in safety (i.e., absence of hazardous situations) while maintaining availability (i.e., ability to perform the mission) compared to state-of-the-art runtime ML monitors in the evaluated dataset.
Evaluating ecological time series is critical for benchmarking model performance in many important applications, including predicting greenhouse gas fluxes, capturing carbon-nitrogen dynamics, and monitoring hydrological cycles. Traditional numerical metrics (e.g., R-squared, root mean square error) have been widely used to quantify the similarity between modeled and observed ecosystem variables, but they often fail to capture domain-specific temporal patterns critical to ecological processes. As a result, these methods are often accompanied by expert visual inspection, which requires substantial human labor and limits the applicability to large-scale evaluation. To address these challenges, we propose a novel framework that integrates metric learning with large language model (LLM)-based natural language policy extraction to develop interpretable evaluation criteria. The proposed method processes pairwise annotations and implements a policy optimization mechanism to generate and combine different assessment metrics. The results obtained on multiple datasets for evaluating the predictions of crop gross primary production and carbon dioxide flux have confirmed the effectiveness of the proposed method in capturing target assessment preferences, including both synthetically generated and expert-annotated model comparisons. The proposed framework bridges the gap between numerical metrics and expert knowledge while providing interpretable evaluation policies that accommodate the diverse needs of different ecosystem modeling studies.
Multi-objective reinforcement learning (MORL) addresses the challenge of simultaneously optimizing multiple, often conflicting, rewards, moving beyond the single-reward focus of conventional reinforcement learning (RL). This approach is essential for applications where agents must balance trade-offs between diverse goals, such as speed, energy efficiency, or stability, as a series of sequential decisions. This paper investigates the applicability and limitations of multi-objective evolutionary algorithms (MOEAs) in solving complex MORL problems. We assess whether these algorithms can effectively address the unique challenges posed by MORL and how MORL instances can serve as benchmarks to evaluate and improve MOEA performance. In particular, we propose a framework to characterize the features influencing MORL instance complexity, select representative MORL problems from the literature, and benchmark a suite of MOEAs alongside single-objective EAs using scalarized MORL formulations. Additionally, we evaluate the utility of existing multi-objective quality indicators in MORL scenarios, such as hypervolume conducting a comparison of the algorithms supported by statistical analysis. Our findings provide insights into the interplay between MORL problem characteristics and algorithmic effectiveness, highlighting opportunities for advancing both MORL research and the design of evolutionary algorithms.
Machine Translation (MT) systems frequently encounter gender-ambiguous occupational terms, where they must assign gender without explicit contextual cues. While individual translations in such cases may not be inherently biased, systematic patterns-such as consistently translating certain professions with specific genders-can emerge, reflecting and perpetuating societal stereotypes. This ambiguity challenges traditional instance-level single-answer evaluation approaches, as no single gold standard translation exists. To address this, we introduce GRAPE, a probability-based metric designed to evaluate gender bias by analyzing aggregated model responses. Alongside this, we present GAMBIT-MT, a benchmarking dataset in English with gender-ambiguous occupational terms. Using GRAPE, we evaluate several MT systems and examine whether their gendered translations in Greek and French align with or diverge from societal stereotypes, real-world occupational gender distributions, and normative standards.
Machine translation (MT) has become indispensable for cross-border communication in globalized industries like e-commerce, finance, and legal services, with recent advancements in large language models (LLMs) significantly enhancing translation quality. However, applying general-purpose MT models to industrial scenarios reveals critical limitations due to domain-specific terminology, cultural nuances, and stylistic conventions absent in generic benchmarks. Existing evaluation frameworks inadequately assess performance in specialized contexts, creating a gap between academic benchmarks and real-world efficacy. To address this, we propose a three-level translation capability framework: (1) Basic Linguistic Competence, (2) Domain-Specific Proficiency, and (3) Cultural Adaptation, emphasizing the need for holistic evaluation across these dimensions. We introduce TransBench, a benchmark tailored for industrial MT, initially targeting international e-commerce with 17,000 professionally translated sentences spanning 4 main scenarios and 33 language pairs. TransBench integrates traditional metrics (BLEU, TER) with Marco-MOS, a domain-specific evaluation model, and provides guidelines for reproducible benchmark construction. Our contributions include: (1) a structured framework for industrial MT evaluation, (2) the first publicly available benchmark for e-commerce translation, (3) novel metrics probing multi-level translation quality, and (4) open-sourced evaluation tools. This work bridges the evaluation gap, enabling researchers and practitioners to systematically assess and enhance MT systems for industry-specific needs.
The challenge of slang translation lies in capturing context-dependent semantic extensions, as slang terms often convey meanings beyond their literal interpretation. While slang detection, explanation, and translation have been studied as isolated tasks in the era of large language models (LLMs), their intrinsic interdependence remains underexplored. The main reason is lacking of a benchmark where the two tasks can be a prerequisite for the third one, which can facilitate idiomatic translation. In this paper, we introduce the interpretative slang translation task (named SlangDIT) consisting of three sub-tasks: slang detection, cross-lingual slang explanation, and slang translation within the current context, aiming to generate more accurate translation with the help of slang detection and slang explanation. To this end, we construct a SlangDIT dataset, containing over 25k English-Chinese sentence pairs. Each source sentence mentions at least one slang term and is labeled with corresponding cross-lingual slang explanation. Based on the benchmark, we propose a deep thinking model, named SlangOWL. It firstly identifies whether the sentence contains a slang, and then judges whether the slang is polysemous and analyze its possible meaning. Further, the SlangOWL provides the best explanation of the slang term targeting on the current context. Finally, according to the whole thought, the SlangOWL offers a suitable translation. Our experiments on LLMs (\emph{e.g.}, Qwen2.5 and LLama-3.1), show that our deep thinking approach indeed enhances the performance of LLMs where the proposed SLangOWL significantly surpasses the vanilla models and supervised fine-tuned models without thinking.
This study explores Large Language Models (LLMs) as autonomous agents for real-world tasks, including freelance software development. This work presents a new benchmark that evaluates LLMs on freelance programming and data analysis tasks derived from economic data. We construct the benchmark using synthetic tasks created from a Kaggle Freelancer dataset of job postings, with all job prices standardized to USD (median fixed-project price around $250, and an average of $306). Each task is accompanied by structured input-output test cases and an estimated price tag, enabling automated correctness checking and a monetary performance valuation. This approach is inspired by OpenAI's recent SWE-Lancer benchmark (1,400 real Upwork tasks worth $1M total). Still, our framework simplifies evaluation using programmatically testable tasks and predicted price values, making it highly scalable and repeatable. On this benchmark, we evaluate four modern LLMs - Claude 3.5 Haiku, GPT-4o-mini, Qwen 2.5, and Mistral. We report each model's accuracy (task success rate and test-case pass rate) and the total "freelance earnings" it achieves (sum of prices of solved tasks). Our results show that Claude 3.5 Haiku performs best, earning approximately $1.52 million USD, followed closely by GPT-4o-mini at $1.49 million, then Qwen 2.5 ($1.33M) and Mistral ($0.70M). We analyze the distribution of errors per task and observe that the strongest models solve the most tasks and rarely fail completely on any project. We discuss the implications of these results for the feasibility of AI as a freelance developer, the advantages and limitations of our automated benchmark approach, and the gap between performance on structured tasks versus the true complexity of real-world freelance jobs.
This study investigates a specific form of positional bias, termed the Myopic Trap, where retrieval models disproportionately attend to the early parts of documents while overlooking relevant information that appears later. To systematically quantify this phenomenon, we propose a semantics-preserving evaluation framework that repurposes the existing NLP datasets into position-aware retrieval benchmarks. By evaluating the SOTA models of full retrieval pipeline, including BM25, embedding models, ColBERT-style late-interaction models, and reranker models, we offer a broader empirical perspective on positional bias than prior work. Experimental results show that embedding models and ColBERT-style models exhibit significant performance degradation when query-related content is shifted toward later positions, indicating a pronounced head bias. Notably, under the same training configuration, ColBERT-style approach show greater potential for mitigating positional bias compared to the traditional single-vector approach. In contrast, BM25 and reranker models remain largely unaffected by such perturbations, underscoring their robustness to positional bias. Code and data are publicly available at: www.github.com/NovaSearch-Team/RAG-Retrieval.