Daily Papers

As large language models (LLMs) become increasingly prevalent and integrated into autonomous systems, ensuring their safety is imperative. Despite significant strides toward safety alignment, recent work GCG~zou2023universal proposes a discrete token optimization algorithm and selects the single suffix with the lowest loss to successfully jailbreak aligned LLMs. In this work, we first discuss the drawbacks of solely picking the suffix with the lowest loss during GCG optimization for jailbreaking and uncover the missed successful suffixes during the intermediate steps. Moreover, we utilize those successful suffixes as training data to learn a generative model, named AmpleGCG, which captures the distribution of adversarial suffixes given a harmful query and enables the rapid generation of hundreds of suffixes for any harmful queries in seconds. AmpleGCG achieves near 100\% attack success rate (ASR) on two aligned LLMs (Llama-2-7B-chat and Vicuna-7B), surpassing two strongest attack baselines. More interestingly, AmpleGCG also transfers seamlessly to attack different models, including closed-source LLMs, achieving a 99\% ASR on the latest GPT-3.5. To summarize, our work amplifies the impact of GCG by training a generative model of adversarial suffixes that is universal to any harmful queries and transferable from attacking open-source LLMs to closed-source LLMs. In addition, it can generate 200 adversarial suffixes for one harmful query in only 4 seconds, rendering it more challenging to defend.

Recent studies demonstrate that Large Language Models (LLMs) are vulnerable to attacks that generate harmful or sensitive outputs. As open-source LLMs are increasingly adopted in high-impact applications such as finance, law, and healthcare, systematically investigating their security risks is becoming increasingly important towards trustworthy LLM era. This paper comprehensively studies effective prompt injection attacks against 14 widely used open-source and three closed-source LLMs on five attack benchmarks. Moreover, existing evaluation metrics mostly only consider the attack success rate, overlooking uncertainty in model responses. Our proposed Attack Success Probability (ASP) additionally captures uncertain behaviors for evaluation, where the model may initially refuse a harmful request but subsequently provide harmful guidance or vice versa, reflecting inconsistency and ambiguity in attack feasibility. By systematically analyzing the effectiveness of prompt injection attacks, we propose a straightforward and effective hypnotism attack; results show that this attack causes aligned language models, including Stablelm2, Mistral, Openchat, and Vicuna, to generate objectionable behaviors, achieving around 90% ASP. They also indicate that ignore prefix attacks can break all 14 open-source LLMs, achieving over 60% ASP on a multi-categorical dataset. We find that moderately well-known LLMs exhibit higher vulnerability to prompt injection attacks, highlighting the need to raise public awareness and prioritize efficient mitigation strategies.

Agentic systems built on large language models (LLMs) offer promising capabilities for automating complex workflows in healthcare AI. We introduce mAIstro, an open-source, autonomous multi-agentic framework for end-to-end development and deployment of medical AI models. The system orchestrates exploratory data analysis, radiomic feature extraction, image segmentation, classification, and regression through a natural language interface, requiring no coding from the user. Built on a modular architecture, mAIstro supports both open- and closed-source LLMs, and was evaluated using a large and diverse set of prompts across 16 open-source datasets, covering a wide range of imaging modalities, anatomical regions, and data types. The agents successfully executed all tasks, producing interpretable outputs and validated models. This work presents the first agentic framework capable of unifying data analysis, AI model development, and inference across varied healthcare applications, offering a reproducible and extensible foundation for clinical and research AI integration. The code is available at: https://github.com/eltzanis/mAIstro

Persona agents, which are LLM agents that act according to an assigned persona, have demonstrated impressive contextual response capabilities across various applications. These persona agents offer significant enhancements across diverse sectors, such as education, healthcare, and entertainment, where model developers can align agent responses to different user requirements thereby broadening the scope of agent applications. However, evaluating persona agent performance is incredibly challenging due to the complexity of assessing persona adherence in free-form interactions across various environments that are relevant to each persona agent. We introduce PersonaGym, the first dynamic evaluation framework for assessing persona agents, and PersonaScore, the first automated human-aligned metric grounded in decision theory for comprehensive large-scale evaluation of persona agents. Our evaluation of 6 open and closed-source LLMs, using a benchmark encompassing 200 personas and 10,000 questions, reveals significant opportunities for advancement in persona agent capabilities across state-of-the-art models. For example, Claude 3.5 Sonnet only has a 2.97% relative improvement in PersonaScore than GPT 3.5 despite being a much more advanced model. Importantly, we find that increased model size and complexity do not necessarily imply enhanced persona agent capabilities thereby highlighting the pressing need for algorithmic and architectural invention towards faithful and performant persona agents.

To solve complex tasks, large language models (LLMs) often require multiple rounds of interactions with the user, sometimes assisted by external tools. However, current evaluation protocols often emphasize benchmark performance with single-turn exchanges, neglecting the nuanced interactions among the user, LLMs, and external tools, while also underestimating the importance of natural language feedback from users. These oversights contribute to discrepancies between research benchmark evaluations and real-world use cases. We introduce MINT, a benchmark that evaluates LLMs' ability to solve tasks with multi-turn interactions by (1) using tools and (2) leveraging natural language feedback. To ensure reproducibility, we provide an evaluation framework where LLMs can access tools by executing Python code and receive users' natural language feedback simulated by GPT-4. We repurpose a diverse set of established evaluation datasets focusing on reasoning, coding, and decision-making and carefully curate them into a compact subset for efficient evaluation. Our analysis of 20 open- and closed-source LLMs offers intriguing findings. (a) LLMs generally benefit from tools and language feedback, with performance gains (absolute, same below) of 1-8% for each turn of tool use and 2-17% with natural language feedback. (b) Better single-turn performance does not guarantee better multi-turn performance. (c) Surprisingly, on the LLMs evaluated, supervised instruction-finetuning (SIFT) and reinforcement learning from human feedback (RLHF) generally hurt multi-turn capabilities. We expect MINT can help measure progress and incentivize research in improving LLMs' capabilities in multi-turn interactions, especially for open-source communities where multi-turn human evaluation can be less accessible compared to commercial LLMs with a larger user base.

Large Language Models (LLMs) have achieved remarkable performance across a wide range of mathematical benchmarks. However, concerns remain as to whether these successes reflect genuine reasoning or superficial pattern recognition. Existing evaluation methods, which typically focus either on the final answer or on the intermediate reasoning steps, reduce mathematical reasoning to a shallow input-output mapping, overlooking its inherently multi-stage and multi-dimensional cognitive nature. Inspired by Polya's problem-solving theory, we propose SMART, a benchmark that decomposes mathematical problem-solving into four cognitive dimensions: Semantic Understanding, Mathematical Reasoning, Arithmetic Computation, and Reflection & Refinement, and introduces dimension-specific tasks to measure the corresponding cognitive processes of LLMs. We apply SMART to 22 state-of-the-art open- and closed-source LLMs and uncover substantial discrepancies in their capabilities across dimensions. Our findings reveal genuine weaknesses in current models and motivate a new metric, the All-Pass Score, designed to better capture true problem-solving capability.

Large Language Models (LLMs) are increasingly capable of generating complete applications from natural language instructions, creating new opportunities in science and education. In these domains, interactive scientific demonstrations are particularly valuable for explaining concepts, supporting new teaching methods, and presenting research findings. Generating such demonstrations requires models to combine accurate scientific knowledge with the ability to implement interactive front-end code that behaves correctly and responds to user actions. This capability goes beyond the scope of existing benchmarks, which typically evaluate either knowledge question answering without grounding in code or static web code generation without scientific interactivity. To evaluate this integrated ability, we design a hybrid framework that combines programmatic functional testing to rigorously verify interaction logic with visually-grounded qualitative testing to assess rendered outputs against reference snapshots. Building on this framework, we present InteractScience, a benchmark consisting of a substantial set of carefully designed questions across five scientific domains, each paired with unit tests, reference snapshots, and checklists. We evaluate 30 leading open- and closed-source LLMs and report results that highlight ongoing weaknesses in integrating domain knowledge with interactive front-end coding. Our work positions InteractScience as the first benchmark to automatically measure this combined capability with realistic interactive operations, providing a foundation for advancing reliable and educationally useful scientific demonstration code generation. All code and data are publicly available at https://github.com/open-compass/InteractScience.

LLM agents have incredible potential for scientific discovery applications. However, the performance of LLM agents on real-world, small molecule drug design (SMDD) tasks across diverse chemistries and targets is unclear. Current evaluation methods are either ad hoc, too simple for real-world discovery, limited in scale, or restricted to single-turn question answering. In effort to standardize the evaluation of LLM agents on small molecule design, we introduce SMDD-Bench, a challenging, multi-turn, long-horizon agentic benchmark consisting of 502 guaranteed-solvable task instances spanning 5 task types: 2D Pharmacophore Identification, Interaction Point Discovery, Scaffold Hopping, Lead Optimization, and Fragment Assembly. SMDD-Bench tasks span a wide region of chemical space and involve 102 unique protein targets. Completely solving the benchmark would require having strong chemical and biological reasoning and 3D intuition, understanding specialized tool use, and displaying planning expertise over a limited number of oracle calls. We benchmark 7 frontier open and closed source LLMs and find even the most performant LLM, GPT5.4, solves only 40.2\% of tasks. We hope SMDD-Bench provides a standardized testbed to invigorate the field towards training and evaluating LLM agents for fully autonomous computational drug design. We host a public leaderboard at smddbench.com .

Natural Question Answering (QA) datasets play a crucial role in evaluating the capabilities of large language models (LLMs), ensuring their effectiveness in real-world applications. Despite the numerous QA datasets that have been developed, there is a notable lack of region-specific datasets generated by native users in their own languages. This gap hinders the effective benchmarking of LLMs for regional and cultural specificities. Furthermore, it also limits the development of fine-tuned models. In this study, we propose a scalable, language-independent framework, NativQA, to seamlessly construct culturally and regionally aligned QA datasets in native languages, for LLM evaluation and tuning. We demonstrate the efficacy of the proposed framework by designing a multilingual natural QA dataset, \mnqa, consisting of ~64k manually annotated QA pairs in seven languages, ranging from high to extremely low resource, based on queries from native speakers from 9 regions covering 18 topics. We benchmark open- and closed-source LLMs with the MultiNativQA dataset. We also showcase the framework efficacy in constructing fine-tuning data especially for low-resource and dialectally-rich languages. We made both the framework NativQA and MultiNativQA dataset publicly available for the community (https://nativqa.gitlab.io).

LaTeX's precision and flexibility in typesetting have made it the gold standard for the preparation of scientific documentation. Large Language Models (LLMs) present a promising opportunity for researchers to produce publication-ready material using LaTeX with natural language instructions, yet current benchmarks completely lack evaluation of this ability. By introducing TeXpert, our benchmark dataset with natural language prompts for generating LaTeX code focused on components of scientific documents across multiple difficulty levels, we conduct an in-depth analysis of LLM performance in this regard and identify frequent error types. Our evaluation across open and closed-source LLMs highlights multiple key findings: LLMs excelling on standard benchmarks perform poorly in LaTeX generation with a significant accuracy drop-off as the complexity of tasks increases; open-source models like DeepSeek v3 and DeepSeek Coder strongly rival closed-source counterparts in LaTeX tasks; and formatting and package errors are unexpectedly prevalent, suggesting a lack of diverse LaTeX examples in the training datasets of most LLMs. Our dataset, code, and model evaluations are available at https://github.com/knowledge-verse-ai/TeXpert.

We present a comprehensive evaluation of the ability of large language models (LLMs) to process culturally grounded language, specifically to understand and pragmatically use figurative expressions that encode local knowledge and cultural nuance. Using figurative language as a proxy for cultural nuance and local knowledge, we design evaluation tasks for contextual understanding, pragmatic use, and connotation interpretation in Arabic and English. We evaluate 22 open- and closed-source LLMs on Egyptian Arabic idioms, multidialectal Arabic proverbs, and English proverbs. Our results show a consistent hierarchy: the average accuracy for Arabic proverbs is 4.29% lower than for English proverbs, and performance for Egyptian idioms is 10.28% lower than for Arabic proverbs. For the pragmatic use task, accuracy drops by 14.07% relative to understanding, though providing contextual idiomatic sentences improves accuracy by 10.66%. Models also struggle with connotative meaning, reaching at most 85.58% agreement with human annotators on idioms with 100% inter-annotator agreement. These findings demonstrate that figurative language serves as an effective diagnostic for cultural reasoning: while LLMs can often interpret figurative meaning, they face challenges in using it appropriately. To support future research, we release Kinayat, the first dataset of Egyptian Arabic idioms designed for both figurative understanding and pragmatic use evaluation.

Human knowledge is inherently structured and interdependent: mastery of a concept requires prior mastery of its prerequisites, a principle formalized by Knowledge Space Theory (KST). While LLMs achieve strong performance on complex reasoning tasks, it remains unclear whether they exhibit coherent, human-like knowledge structure. We introduce a KST-grounded framework for evaluating LLM knowledge structure in mathematical reasoning, using it as a normative framework to analyze whether LLM behavior adheres to principled knowledge dependencies. Evaluating eight open- and closed-source LLMs against real human learners, we find that (1) LLMs do not adhere to human knowledge structure -- they frequently violate knowledge dependencies and fail to leverage related knowledge provided in context to improve performance on dependent questions; (2) LLMs do not share a consistent knowledge structure among themselves, as reflected by low overlap in their knowledge distributions. Furthermore, these structural deficiencies remain largely invisible to accuracy-based and LLM-as-judge evaluations. Together, our results provide behavioral evidence that current LLMs knowledge does not follow a human-like structure.

Evaluating the occurrence and triggers of large language model (LLM) behaviors - such as sycophancy, self-preference, or over-confidence - is critical for predicting real-world model deployment risks. However, existing situated behavioral evaluations typically ignore cultural context, limiting their generalizability across an increasingly global user base. To address this gap, we propose CuBEs - Culturally-situated Behavior Evaluations that probe for response patterns across diverse user cultures. We first extend an automated testing pipeline to inject cultural context into behavioral test scenarios and subsequent evaluation. We assess the cultural adaptability of this pipeline by building a human-labeled dataset that captures nuanced dimensions of behavior understanding across 12 distinct cultures. Our dataset reveals significant cross-cultural variations that one-size-fits all judgments fail to capture. Through evaluating 13 open- and closed-source LLMs, we find that introducing cultural situatedness in the evaluation scenario creates significant variation in the presence of a behavior. For example, while our baseline experiments testing for political bias capture localized Western political dimensions like the American conservative-progressive divide, non-Western culturally situated evaluations surface entirely different axes of bias such as religious and colonial political issues. Our findings demonstrate that standard, culturally-agnostic evaluations fail to capture these shifts, highlighting the necessity of culturally situated behavioral testing for global deployments.

Emotional intelligence (EI) in Large Language Models (LLMs) is often evaluated through static understanding tasks or single-response dialogue generation. However, emotion management is interactive: a good model should not only recognize a user's emotion, but also improve the user's emotional and relational state over several turns. We introduce EIBench, a simulator-based benchmark for interactive emotion management. EIBench contains 2,222 scenarios, with 2,009 for training and 213 for held-out testing. The scenarios are organized by a 2x2 taxonomy covering Support, Defense, Repair, and Charm, which together capture different forms of support, boundary maintenance, trust repair, and rapport building. In each scenario, an LLM simulator plays the user, updates an emotion-relation state after each turn, and maps the final state to an anchor-based score. This design makes EIBench both an evaluation benchmark and a training environment: the final state gives the outcome reward, while the per-turn state updates provide dense feedback for RL. We evaluate 15 open- and closed-source LLMs. Current models perform well on support and rapport-building scenes, but struggle with boundary maintenance under user pressure. To improve the EI ability of LLMs, we propose Centered Turn-Credit GRPO (CTC-GRPO), a GRPO extension that reuses the simulator's per-turn state updates as dense turn-level feedback while preserving the final outcome reward. CTC-GRPO improves Qwen3-8B from -22.4 to +22.4 on EIBench and also improves on out-of-distribution evaluations including SAGE (+12.4) and EQBench3 (+20.9%). Our results show that simulator-tracked user states can support both evaluation and training for multi-turn emotion management.

Tool calling allows large language models (LLMs) to interact with external systems like APIs, enabling applications in customer support, data analysis, and dynamic content generation. While recent benchmarks have advanced tool-use research, they suffer from key limitations, including reliance on simulated or restricted APIs, limited reproducibility, and a lack of cultural and geographic diversity. To address these gaps, we introduce International Tool Calling (ITC), a large-scale, multilingual benchmark designed for realistic, globally distributed tool-calling scenarios. ITC includes 3,571 real APIs and 17,540 tool calling tasks across 20 categories and 40 countries. Experiments reveal substantial performance gaps between open- and closed-source LLMs, while fine-tuning on ITC yields significant improvements, particularly for non-English queries, enhancing cross-lingual generalization, reasoning consistency, and robustness to out-of-domain tools. ITC provides a valuable benchmark for advancing LLM robustness and performance in complex, multi-tool, and international scenarios. Dataset: https://anonymous.4open.science/r/International-Tool-Calling-ITC-dataset-FAF4/.

This paper presents a case study on the design, administration, post-processing, and evaluation of surveys on large language models (LLMs). It comprises two components: (1) A statistical method for eliciting beliefs encoded in LLMs. We introduce statistical measures and evaluation metrics that quantify the probability of an LLM "making a choice", the associated uncertainty, and the consistency of that choice. (2) We apply this method to study what moral beliefs are encoded in different LLMs, especially in ambiguous cases where the right choice is not obvious. We design a large-scale survey comprising 680 high-ambiguity moral scenarios (e.g., "Should I tell a white lie?") and 687 low-ambiguity moral scenarios (e.g., "Should I stop for a pedestrian on the road?"). Each scenario includes a description, two possible actions, and auxiliary labels indicating violated rules (e.g., "do not kill"). We administer the survey to 28 open- and closed-source LLMs. We find that (a) in unambiguous scenarios, most models "choose" actions that align with commonsense. In ambiguous cases, most models express uncertainty. (b) Some models are uncertain about choosing the commonsense action because their responses are sensitive to the question-wording. (c) Some models reflect clear preferences in ambiguous scenarios. Specifically, closed-source models tend to agree with each other.

As Large Language Models (LLMs) become deeply integrated into human life and increasingly influence decision-making, it's crucial to evaluate whether and to what extent they exhibit subjective preferences, opinions, and beliefs. These tendencies may stem from biases within the models, which may shape their behavior, influence the advice and recommendations they offer to users, and potentially reinforce certain viewpoints. This paper presents the Preference, Opinion, and Belief survey (POBs), a benchmark developed to assess LLMs' subjective inclinations across societal, cultural, ethical, and personal domains. We applied our benchmark to evaluate leading open- and closed-source LLMs, measuring desired properties such as reliability, neutrality, and consistency. In addition, we investigated the effect of increasing the test-time compute, through reasoning and self-reflection mechanisms, on those metrics. While effective in other tasks, our results show that these mechanisms offer only limited gains in our domain. Furthermore, we reveal that newer model versions are becoming less consistent and more biased toward specific viewpoints, highlighting a blind spot and a concerning trend. POBS: https://ibm.github.io/POBS

Despite the strong performance of large language models (LLMs) across a wide range of tasks, they still have reliability issues. Previous studies indicate that strong LLMs like GPT-4-turbo excel in evaluating the reliability of responses from LLMs, but face efficiency and local deployment issues. Thus, to enable weak LLMs to effectively assess the reliability of LLM responses, we propose a novel cross-query-comparison-based method called Meta Ranking (MR). Unlike previous few-shot methods that solely based on in-context learning capabilities in LLMs, MR assesses reliability by pairwisely ranking the target query-response pair with multiple reference query-response pairs. We found that MR is highly effective in error detection for LLM responses, where weak LLMs, such as Phi-2, could surpass strong baselines like GPT-3.5-turbo, requiring only five reference samples and significantly improving efficiency. We further demonstrate that MR can enhance strong LLMs' performance in two practical applications: model cascading and instruction tuning. In model cascading, we combine open- and closed-source LLMs to achieve performance comparable to GPT-4-turbo with lower costs. In instruction tuning, we use MR for iterative training data filtering, significantly reducing data processing time and enabling LLaMA-7B and Phi-2 to surpass Alpaca-13B with fewer training tokens. These results underscore the high potential of MR in both efficiency and effectiveness.

Although Large Language Models (LLMs) achieve remarkable performance across various tasks, they often struggle with complex reasoning tasks, such as answering mathematical questions. Recent efforts to address this issue have primarily focused on leveraging mathematical datasets through supervised fine-tuning or self-improvement techniques. However, these methods often depend on high-quality datasets that are difficult to prepare, or they require substantial computational resources for fine-tuning. Inspired by findings that LLMs know how to produce the right answer but struggle to select the correct reasoning path, we propose a purely inference-based searching method -- MindStar (M*). This method formulates reasoning tasks as searching problems and proposes two search ideas to identify the optimal reasoning paths. We evaluate the M* framework on both the GSM8K and MATH datasets, comparing its performance with existing open and closed-source LLMs. Our results demonstrate that M* significantly enhances the reasoning abilities of open-source models, such as Llama-2-13B and Mistral-7B, and achieves comparable performance to GPT-3.5 and Grok-1, but with substantially reduced model size and computational costs.

Recent work on large language models (LLMs) has increasingly focused on post-training and alignment with datasets curated to enhance instruction following, world knowledge, and specialized skills. However, most post-training datasets used in leading open- and closed-source LLMs remain inaccessible to the public, with limited information about their construction process. This lack of transparency has motivated the recent development of open-source post-training corpora. While training on these open alternatives can yield performance comparable to that of leading models, systematic comparisons remain challenging due to the significant computational cost of conducting them rigorously at scale, and are therefore largely absent. As a result, it remains unclear how specific samples, task types, or curation strategies influence downstream performance when assessing data quality. In this work, we conduct the first comprehensive side-by-side analysis of two prominent open post-training datasets: Tulu-3-SFT-Mix and SmolTalk. Using the Magpie framework, we annotate each sample with detailed quality metrics, including turn structure (single-turn vs. multi-turn), task category, input quality, and response quality, and we derive statistics that reveal structural and qualitative similarities and differences between the two datasets. Based on these insights, we design a principled curation recipe that produces a new data mixture, TuluTalk, which contains 14% fewer samples than either source dataset while matching or exceeding their performance on key benchmarks. Our findings offer actionable insights for constructing more effective post-training datasets that improve model performance within practical resource limits. To support future research, we publicly release both the annotated source datasets and our curated TuluTalk mixture.

Large language models (LLMs) have shown remarkable abilities in different fields, including standard Natural Language Processing (NLP) tasks. To elicit knowledge from LLMs, prompts play a key role, consisting of natural language instructions. Most open and closed source LLMs are trained on available labeled and unlabeled resources--digital content such as text, images, audio, and videos. Hence, these models have better knowledge for high-resourced languages but struggle with low-resourced languages. Since prompts play a crucial role in understanding their capabilities, the language used for prompts remains an important research question. Although there has been significant research in this area, it is still limited, and less has been explored for medium to low-resourced languages. In this study, we investigate different prompting strategies (native vs. non-native) on 11 different NLP tasks associated with 12 different Arabic datasets (9.7K data points). In total, we conducted 197 experiments involving 3 LLMs, 12 datasets, and 3 prompting strategies. Our findings suggest that, on average, the non-native prompt performs the best, followed by mixed and native prompts.

Large language models (LLMs) have achieved impressive success on many benchmarks for mathematical reasoning. However, there is growing concern that some of this performance actually reflects dataset contamination, where data closely resembling benchmark questions leaks into the training data, instead of true reasoning ability. To investigate this claim rigorously, we commission Grade School Math 1000 (GSM1k). GSM1k is designed to mirror the style and complexity of the established GSM8k benchmark, the gold standard for measuring elementary mathematical reasoning. We ensure that the two benchmarks are comparable across important metrics such as human solve rates, number of steps in solution, answer magnitude, and more. When evaluating leading open- and closed-source LLMs on GSM1k, we observe accuracy drops of up to 13%, with several families of models (e.g., Phi and Mistral) showing evidence of systematic overfitting across almost all model sizes. At the same time, many models, especially those on the frontier, (e.g., Gemini/GPT/Claude) show minimal signs of overfitting. Further analysis suggests a positive relationship (Spearman's r^2=0.32) between a model's probability of generating an example from GSM8k and its performance gap between GSM8k and GSM1k, suggesting that many models may have partially memorized GSM8k.

Advanced applied mathematics problems are underrepresented in existing Large Language Model (LLM) benchmark datasets. To address this, we introduce HARDMath, a dataset inspired by a graduate course on asymptotic methods, featuring challenging applied mathematics problems that require analytical approximation techniques. These problems demand a combination of mathematical reasoning, computational tools, and subjective judgment, making them difficult for LLMs. Our framework auto-generates a large number of problems with solutions validated against numerical ground truths. We evaluate both open- and closed-source LLMs on HARDMath-mini, a sub-sampled test set of 366 problems, as well as on 40 word problems formulated in applied science contexts. Even leading closed-source models like GPT-4 achieve only 43.8% overall accuracy with few-shot Chain-of-Thought prompting, and all models demonstrate significantly lower performance compared to results on existing mathematics benchmark datasets. We additionally conduct a detailed error analysis to gain insights into the failure cases of LLMs. These results demonstrate limitations of current LLM performance on advanced graduate-level applied math problems and underscore the importance of datasets like HARDMath to advance mathematical abilities of LLMs.

Large Language Models (LLMs)-based agents have made impressive progress in reasoning and tool use, enabling them to solve complex tasks. However, their ability to proactively collaborate with users, especially when goals are vague, evolving, or indirectly expressed, remains underexplored. To address this gap, we introduce UserBench, a user-centric benchmark designed to evaluate agents in multi-turn, preference-driven interactions. UserBench features simulated users who start with underspecified goals and reveal preferences incrementally, requiring agents to proactively clarify intent and make grounded decisions with tools. Our evaluation of leading open- and closed-source LLMs reveals a significant disconnect between task completion and user alignment. For instance, models provide answers that fully align with all user intents only 20% of the time on average, and even the most advanced models uncover fewer than 30% of all user preferences through active interaction. These results highlight the challenges of building agents that are not just capable task executors, but true collaborative partners. UserBench offers an interactive environment to measure and advance this critical capability.

Large Language Models (LLMs) are commonly trained on multilingual corpora that include Greek, yet reliable evaluation benchmarks for Greek-particularly those based on authentic, native-sourced content-remain limited. Existing datasets are often machine-translated from English, failing to capture Greek linguistic and cultural characteristics. We introduce GreekMMLU, a native-sourced benchmark for massive multitask language understanding in Greek, comprising 21,805 multiple-choice questions across 45 subject areas, organized under a newly defined subject taxonomy and annotated with educational difficulty levels spanning primary to professional examinations. All questions are sourced or authored in Greek from academic, professional, and governmental exams. We publicly release 16,857 samples and reserve 4,948 samples for a private leaderboard to enable robust and contamination-resistant evaluation. Evaluations of over 80 open- and closed-source LLMs reveal substantial performance gaps between frontier and open-weight models, as well as between Greek-adapted models and general multilingual ones. Finally, we provide a systematic analysis of factors influencing performance-including model scale, adaptation, and prompting-and derive insights for improving LLM capabilities in Greek.

Evaluating large language models (LLMs) across languages remains challenging, as most multilingual benchmarks rely on translated English datasets, often obscuring linguistic and cultural specificity in the target language. This issue is particularly pronounced for less-resourced languages such as Kyrgyz, where reliable natively authored evaluation data are scarce. Building on previously introduced Kyrgyz-language evaluation datasets, this work reports the first systematic and large-scale evaluation of LLMs in Kyrgyz using the KyrgyzLLM-Bench benchmark suite. KyrgyzLLM-Bench comprises two natively authored datasets-KyrgyzMMLU and KyrgyzRC-together with carefully translated and manually post-edited versions of WinoGrande, HellaSwag, BoolQ, and TruthfulQA. We evaluate 26 open- and closed-source LLMs under zero-shot and few-shot settings, analyzing model performance, cross-lingual transfer, and the impact of translation artifacts on evaluation reliability. Across families and tasks, model rankings transfer broadly from English to Kyrgyz on WinoGrande and BoolQ, and to a lesser extent on MMLU, while HellaSwag exhibits a substantial English-Kyrgyz performance gap consistent with translation-induced plausibility shifts. Few-shot prompting improves several open-source models on reading comprehension but behaves inconsistently for proprietary models on translated tasks. We publicly release all datasets, evaluation code, and per-model results, and integrate the Kyrgyz tasks into a widely used multilingual evaluation framework to support future research on Kyrgyz NLP.

Large language models (LLMs) are increasingly used as assistants for statistical and data science work, yet existing evaluations largely assume the analysis target is already specified. In practice, users arrive with informal goals and heterogeneous data, leaving the model to decide what statistical task is implied and which data are relevant. We first formalize this upstream step as Statistical Problem Formulation and decompose it into two subtasks: (1) Statistical Problem Classification and (2) Variable Identification & Role Assignment. We then introduce StatFormBench, a benchmark built from five cross-domain statistics textbooks and a data science case library, covering diverse problem types, data representations, and scenario styles. It contains 1,013 samples spanning 20 coarse-grained and 85 fine-grained statistical problem categories. Across 14 open- and closed-source LLMs, the best zero-shot models reach only 72.0 fine-grained classification accuracy and 63.2 variable set overlap. No model performs consistently best across the two subtasks, while enhanced prompting strategies yield only limited or inconsistent gains. We release the benchmark data on Hugging Face at https://huggingface.co/datasets/THU-CongLab/StatFormBench and the evaluation code on GitHub at https://github.com/THU-CongLab/StatFormBench.

Agents based on Large Language Models (LLMs) are increasingly permeating various domains of human production and life, highlighting the importance of aligning them with human values. The current alignment of AI systems primarily focuses on passively aligning LLMs through human intervention. However, agents possess characteristics like receiving environmental feedback and self-evolution, rendering the LLM alignment methods inadequate. In response, we propose an evolutionary framework for agent evolution and alignment, named EvolutionaryAgent, which transforms agent alignment into a process of evolution and selection under the principle of survival of the fittest. In an environment where social norms continuously evolve, agents better adapted to the current social norms will have a higher probability of survival and proliferation, while those inadequately aligned dwindle over time. Experimental results assessing the agents from multiple perspectives in aligning with social norms demonstrate that EvolutionaryAgent can align progressively better with the evolving social norms while maintaining its proficiency in general tasks. Effectiveness tests conducted on various open and closed-source LLMs as the foundation for agents also prove the applicability of our approach.

Data analytics is essential for extracting valuable insights from data that can assist organizations in making effective decisions. We introduce InsightBench, a benchmark dataset with three key features. First, it consists of 100 datasets representing diverse business use cases such as finance and incident management, each accompanied by a carefully curated set of insights planted in the datasets. Second, unlike existing benchmarks focusing on answering single queries, InsightBench evaluates agents based on their ability to perform end-to-end data analytics, including formulating questions, interpreting answers, and generating a summary of insights and actionable steps. Third, we conducted comprehensive quality assurance to ensure that each dataset in the benchmark had clear goals and included relevant and meaningful questions and analysis. Furthermore, we implement a two-way evaluation mechanism using LLaMA-3 as an effective, open-source evaluator to assess agents' ability to extract insights. We also propose AgentPoirot, our baseline data analysis agent capable of performing end-to-end data analytics. Our evaluation on InsightBench shows that AgentPoirot outperforms existing approaches (such as Pandas Agent) that focus on resolving single queries. We also compare the performance of open- and closed-source LLMs and various evaluation strategies. Overall, this benchmark serves as a testbed to motivate further development in comprehensive automated data analytics and can be accessed here: https://github.com/ServiceNow/insight-bench.

Large language models (LLMs) have significantly transformed the educational landscape. As current plagiarism detection tools struggle to keep pace with LLMs' rapid advancements, the educational community faces the challenge of assessing students' true problem-solving abilities in the presence of LLMs. In this work, we explore a new paradigm for ensuring fair evaluation -- generating adversarial examples which preserve the structure and difficulty of the original questions aimed for assessment, but are unsolvable by LLMs. Focusing on the domain of math word problems, we leverage abstract syntax trees to structurally generate adversarial examples that cause LLMs to produce incorrect answers by simply editing the numeric values in the problems. We conduct experiments on various open- and closed-source LLMs, quantitatively and qualitatively demonstrating that our method significantly degrades their math problem-solving ability. We identify shared vulnerabilities among LLMs and propose a cost-effective approach to attack high-cost models. Additionally, we conduct automatic analysis to investigate the cause of failure, providing further insights into the limitations of LLMs.

Biomedical researchers increasingly rely on large-scale structured databases for complex analytical tasks. However, current text-to-SQL systems often struggle to map qualitative scientific questions into executable SQL, particularly when implicit domain reasoning is required. We introduce BiomedSQL, the first benchmark explicitly designed to evaluate scientific reasoning in text-to-SQL generation over a real-world biomedical knowledge base. BiomedSQL comprises 68,000 question/SQL query/answer triples grounded in a harmonized BigQuery knowledge base that integrates gene-disease associations, causal inference from omics data, and drug approval records. Each question requires models to infer domain-specific criteria, such as genome-wide significance thresholds, effect directionality, or trial phase filtering, rather than rely on syntactic translation alone. We evaluate a range of open- and closed-source LLMs across prompting strategies and interaction paradigms. Our results reveal a substantial performance gap: GPT-o3-mini achieves 59.0% execution accuracy, while our custom multi-step agent, BMSQL, reaches 62.6%, both well below the expert baseline of 90.0%. BiomedSQL provides a new foundation for advancing text-to-SQL systems capable of supporting scientific discovery through robust reasoning over structured biomedical knowledge bases. Our dataset is publicly available at https://huggingface.co/datasets/NIH-CARD/BiomedSQL, and our code is open-source at https://github.com/NIH-CARD/biomedsql.

The ability of Large Language Models (LLMs) to generate factually correct output remains relatively unexplored due to the lack of fact-checking and knowledge grounding during training and inference. In this work, we aim to address this challenge through the Entity Disambiguation (ED) task. We first consider prompt engineering, and design a three-step hard-prompting method to probe LLMs' ED performance without supervised fine-tuning (SFT). Overall, the prompting method improves the micro-F_1 score of the original vanilla models by a large margin, on some cases up to 36% and higher, and obtains comparable performance across 10 datasets when compared to existing methods with SFT. We further improve the knowledge grounding ability through instruction tuning (IT) with similar prompts and responses. The instruction-tuned model not only achieves higher micro-F1 score performance as compared to several baseline methods on supervised entity disambiguation tasks with an average micro-F_1 improvement of 2.1% over the existing baseline models, but also obtains higher accuracy on six Question Answering (QA) tasks in the zero-shot setting. Our methodologies apply to both open- and closed-source LLMs.

There is growing interest in ensuring that large language models (LLMs) align with human values. However, the alignment of such models is vulnerable to adversarial jailbreaks, which coax LLMs into overriding their safety guardrails. The identification of these vulnerabilities is therefore instrumental in understanding inherent weaknesses and preventing future misuse. To this end, we propose Prompt Automatic Iterative Refinement (PAIR), an algorithm that generates semantic jailbreaks with only black-box access to an LLM. PAIR -- which is inspired by social engineering attacks -- uses an attacker LLM to automatically generate jailbreaks for a separate targeted LLM without human intervention. In this way, the attacker LLM iteratively queries the target LLM to update and refine a candidate jailbreak. Empirically, PAIR often requires fewer than twenty queries to produce a jailbreak, which is orders of magnitude more efficient than existing algorithms. PAIR also achieves competitive jailbreaking success rates and transferability on open and closed-source LLMs, including GPT-3.5/4, Vicuna, and PaLM-2.

Recent progress in large language models (LLMs) has enabled tool-augmented agents capable of solving complex real-world tasks through step-by-step reasoning. However, existing evaluations often focus on general-purpose or multimodal scenarios, leaving a gap in domain-specific benchmarks that assess tool-use capabilities in complex remote sensing use cases. We present ThinkGeo, an agentic benchmark designed to evaluate LLM-driven agents on remote sensing tasks via structured tool use and multi-step planning. Inspired by tool-interaction paradigms, ThinkGeo includes human-curated queries spanning a wide range of real-world applications such as urban planning, disaster assessment and change analysis, environmental monitoring, transportation analysis, aviation monitoring, recreational infrastructure, and industrial site analysis. Queries are grounded in satellite or aerial imagery, including both optical RGB and SAR data, and require agents to reason through a diverse toolset. We implement a ReAct-style interaction loop and evaluate both open and closed-source LLMs (e.g., GPT-4o, Qwen2.5) on 486 structured agentic tasks with 1,778 expert-verified reasoning steps. The benchmark reports both step-wise execution metrics and final answer correctness. Our analysis reveals notable disparities in tool accuracy and planning consistency across models. ThinkGeo provides the first extensive testbed for evaluating how tool-enabled LLMs handle spatial reasoning in remote sensing.

Large Language Models (LLMs) have recently demonstrated strong capabilities in tool use, yet progress in tool retrieval remains hindered by incomplete and heterogeneous tool documentation. To address this challenge, we introduce Tool-DE, a new benchmark and framework that systematically enriches tool documentation with structured fields to enable more effective tool retrieval, together with two dedicated models, Tool-Embed and Tool-Rank. We design a scalable document expansion pipeline that leverages both open- and closed-source LLMs to generate, validate, and refine enriched tool profiles at low cost, producing large-scale corpora with 50k instances for embedding-based retrievers and 200k for rerankers. On top of this data, we develop two models specifically tailored for tool retrieval: Tool-Embed, a dense retriever, and Tool-Rank, an LLM-based reranker. Extensive experiments on ToolRet and Tool-DE demonstrate that document expansion substantially improves retrieval performance, with Tool-Embed and Tool-Rank achieving new state-of-the-art results on both benchmarks. We further analyze the contribution of individual fields to retrieval effectiveness, as well as the broader impact of document expansion on both training and evaluation. Overall, our findings highlight both the promise and limitations of LLM-driven document expansion, positioning Tool-DE, along with the proposed Tool-Embed and Tool-Rank, as a foundation for future research in tool retrieval.

Large language models (LLMs) can generate chains of thought (CoTs) that are not always causally responsible for their final outputs. When such a mismatch occurs, the CoT no longer faithfully reflects the decision-critical factors driving the model's behavior, leading to the reduced CoT monitorability problem. However, a comprehensive and fully open-source benchmark for studying CoT monitorability remains lacking. To address this gap, we propose MonitorBench, a systematic benchmark for evaluating CoT monitorability in LLMs. MonitorBench provides: (1) a diverse set of 1,514 test instances with carefully designed decision-critical factors across 19 tasks spanning 7 categories to characterize when CoTs can be used to monitor the factors driving LLM behavior; and (2) two stress-test settings to quantify the extent to which CoT monitorability can be degraded. Extensive experiments across multiple popular LLMs with varying capabilities show that CoT monitorability is higher when producing the final target response requires structural reasoning through the decision-critical factor. Closed-source LLMs generally show lower monitorability, and there exists a negative relationship between monitorability and model capability. Moreover, both open- and closed-source LLMs can intentionally reduce monitorability under stress-tests, with monitorability dropping by up to 30% in some tasks that do not require structural reasoning over the decision-critical factors. Beyond these empirical insights, MonitorBench provides a basis for further research on evaluating future LLMs, studying advanced stress-test monitorability techniques, and developing new monitoring approaches.

Large language models (LLMs) are typically aligned to a universal set of safety and usage principles intended for broad public acceptability. Yet, real-world applications of LLMs often take place within organizational ecosystems shaped by distinctive corporate policies, regulatory requirements, use cases, brand guidelines, and ethical commitments. This reality highlights the need for rigorous and comprehensive evaluation of LLMs with pluralistic alignment goals, an alignment paradigm that emphasizes adaptability to diverse user values and needs. In this work, we present PLURALISTIC BEHAVIOR SUITE (PBSUITE), a dynamic evaluation suite designed to systematically assess LLMs' capacity to adhere to pluralistic alignment specifications in multi-turn, interactive conversations. PBSUITE consists of (1) a diverse dataset of 300 realistic LLM behavioral policies, grounded in 30 industries; and (2) a dynamic evaluation framework for stress-testing model compliance with custom behavioral specifications under adversarial conditions. Using PBSUITE, We find that leading open- and closed-source LLMs maintain robust adherence to behavioral policies in single-turn settings (less than 4% failure rates), but their compliance weakens substantially in multi-turn adversarial interactions (up to 84% failure rates). These findings highlight that existing model alignment and safety moderation methods fall short in coherently enforcing pluralistic behavioral policies in real-world LLM interactions. Our work contributes both the dataset and analytical framework to support future research toward robust and context-aware pluralistic alignment techniques.

Large language models (LLMs) achieve remarkable advancements by leveraging tools to interact with external environments, a critical step toward generalized AI. However, the standard supervised fine-tuning (SFT) approach, which relies on large-scale datasets, often overlooks task-specific characteristics in tool use, leading to performance bottlenecks. To address this issue, we analyze three existing LLMs and uncover key insights: training data can inadvertently impede tool-use behavior, token importance is distributed unevenly, and errors in tool calls fall into a small set of distinct categories. Building on these findings, we propose TL-Training, a task-feature-based framework that mitigates the effects of suboptimal training data, dynamically adjusts token weights to prioritize key tokens during SFT, and incorporates a robust reward mechanism tailored to error categories, optimized through proximal policy optimization. We validate TL-Training by training CodeLLaMA-2-7B and evaluating it on four diverse open-source test sets. Our results demonstrate that the LLM trained by our method matches or surpasses both open- and closed-source LLMs in tool-use performance using only 1,217 training data points. Additionally, our method enhances robustness in noisy environments and improves general task performance, offering a scalable and efficient paradigm for tool-use training in LLMs. The code and data are available at https://github.com/Junjie-Ye/TL-Training.

Artificial intelligence (AI) and large language models (LLMs) in healthcare require advanced clinical skills (CS), yet current benchmarks fail to evaluate these comprehensively. We introduce MedQA-CS, an AI-SCE framework inspired by medical education's Objective Structured Clinical Examinations (OSCEs), to address this gap. MedQA-CS evaluates LLMs through two instruction-following tasks, LLM-as-medical-student and LLM-as-CS-examiner, designed to reflect real clinical scenarios. Our contributions include developing MedQA-CS, a comprehensive evaluation framework with publicly available data and expert annotations, and providing the quantitative and qualitative assessment of LLMs as reliable judges in CS evaluation. Our experiments show that MedQA-CS is a more challenging benchmark for evaluating clinical skills than traditional multiple-choice QA benchmarks (e.g., MedQA). Combined with existing benchmarks, MedQA-CS enables a more comprehensive evaluation of LLMs' clinical capabilities for both open- and closed-source LLMs.

Large language models (LLMs) are evolving into agentic systems that reason, plan, and operate external tools. The Model Context Protocol (MCP) is a key enabler of this transition, offering a standardized interface for connecting LLMs with heterogeneous tools and services. Yet MCP's openness and multi-server workflows introduce new safety risks that existing benchmarks fail to capture, as they focus on isolated attacks or lack real-world coverage. We present MCP-SafetyBench, a comprehensive benchmark built on real MCP servers that supports realistic multi-turn evaluation across five domains: browser automation, financial analysis, location navigation, repository management, and web search. It incorporates a unified taxonomy of 20 MCP attack types spanning server, host, and user sides, and includes tasks requiring multi-step reasoning and cross-server coordination under uncertainty. Using MCP-SafetyBench, we systematically evaluate leading open- and closed-source LLMs, revealing that all models remain vulnerable to MCP attacks, with a notable safety-utility trade-off. Our results highlight the urgent need for stronger defenses and establish MCP-SafetyBench as a foundation for diagnosing and mitigating safety risks in real-world MCP deployments.

We study the use of large language model-based agents for interacting with software via web browsers. Unlike prior work, we focus on measuring the agents' ability to perform tasks that span the typical daily work of knowledge workers utilizing enterprise software systems. To this end, we propose WorkArena, a remote-hosted benchmark of 29 tasks based on the widely-used ServiceNow platform. We also introduce BrowserGym, an environment for the design and evaluation of such agents, offering a rich set of actions as well as multimodal observations. Our empirical evaluation reveals that while current agents show promise on WorkArena, there remains a considerable gap towards achieving full task automation. Notably, our analysis uncovers a significant performance disparity between open and closed-source LLMs, highlighting a critical area for future exploration and development in the field.

When building Large Language Models (LLMs), it is paramount to bear safety in mind and protect them with guardrails. Indeed, LLMs should never generate content promoting or normalizing harmful, illegal, or unethical behavior that may contribute to harm to individuals or society. This principle applies to both normal and adversarial use. In response, we introduce ALERT, a large-scale benchmark to assess safety based on a novel fine-grained risk taxonomy. It is designed to evaluate the safety of LLMs through red teaming methodologies and consists of more than 45k instructions categorized using our novel taxonomy. By subjecting LLMs to adversarial testing scenarios, ALERT aims to identify vulnerabilities, inform improvements, and enhance the overall safety of the language models. Furthermore, the fine-grained taxonomy enables researchers to perform an in-depth evaluation that also helps one to assess the alignment with various policies. In our experiments, we extensively evaluate 10 popular open- and closed-source LLMs and demonstrate that many of them still struggle to attain reasonable levels of safety.

While Chain of Thought (CoT) prompting approaches have significantly consolidated the reasoning capabilities of large language models (LLMs), they still face limitations that require extensive human effort or have performance needs to be improved. Existing endeavors have focused on bridging these gaps; however, these approaches either hinge on external data and cannot completely eliminate manual effort, or they fall short in effectively directing LLMs to generate high-quality exemplary prompts. To address the said pitfalls, we propose a novel prompt approach for automatic reasoning named LBS3, inspired by curriculum learning which better reflects human learning habits. Specifically, LBS3 initially steers LLMs to recall easy-to-hard proxy queries that are pertinent to the target query. Following this, it invokes a progressive strategy that utilizes exemplary prompts stemmed from easy-proxy queries to direct LLMs in solving hard-proxy queries, enabling the high-quality of the proxy solutions. Finally, our extensive experiments in various reasoning-intensive tasks with varying open- and closed-source LLMs show that LBS3 achieves strongly competitive performance compared to the SOTA baselines.

In-context learning (ICL) is an important yet not fully understood ability of pre-trained large language models (LLMs). It can greatly enhance task performance using a few examples, termed demonstrations, without fine-tuning. Although effective in question answering, ICL often underperforms in long-form generation tasks such as summarization. Under appropriately realistic assumptions, we empirically and theoretically show that ICL demonstrations alone are insufficient to teach LLMs the task language and format distributions for generation. We argue for explicit exposure to the task distributions and hypothesize that defining them by prompting enhances model performance. To this end, we present LongGuide, which efficiently generates two parallel streams of guidelines capturing task language and format properties: (i) Metric Guidelines (MGs) that instruct models to optimize self-evaluated metrics; and (ii) Output Constraint Guidelines (OCGs) that constrain generation at both token and sentence levels. LongGuide automatically selects the best combination of guidelines, improving both strong open- and closed-source LLMs by over 5% in both zero- and few-shot settings. We show that LongGuide is generalizable, learnable by weak models to enhance strong ones, and integrates synergistically with automatic prompt optimizers.

Large language models (LLMs) have gained widespread interest due to their ability to process human language and perform tasks on which they have not been explicitly trained. This is relevant for the chemical sciences, which face the problem of small and diverse datasets that are frequently in the form of text. LLMs have shown promise in addressing these issues and are increasingly being harnessed to predict chemical properties, optimize reactions, and even design and conduct experiments autonomously. However, we still have only a very limited systematic understanding of the chemical reasoning capabilities of LLMs, which would be required to improve models and mitigate potential harms. Here, we introduce "ChemBench," an automated framework designed to rigorously evaluate the chemical knowledge and reasoning abilities of state-of-the-art LLMs against the expertise of human chemists. We curated more than 7,000 question-answer pairs for a wide array of subfields of the chemical sciences, evaluated leading open and closed-source LLMs, and found that the best models outperformed the best human chemists in our study on average. The models, however, struggle with some chemical reasoning tasks that are easy for human experts and provide overconfident, misleading predictions, such as about chemicals' safety profiles. These findings underscore the dual reality that, although LLMs demonstrate remarkable proficiency in chemical tasks, further research is critical to enhancing their safety and utility in chemical sciences. Our findings also indicate a need for adaptations to chemistry curricula and highlight the importance of continuing to develop evaluation frameworks to improve safe and useful LLMs.

ML libraries, often written in architecture-specific programming languages (ASPLs) that target domain-specific architectures, are key to efficient ML systems. However, writing these high-performance ML libraries is challenging because it requires expert knowledge of ML algorithms and the ASPL. Large language models (LLMs), on the other hand, have shown general coding capabilities. However, challenges remain when using LLMs for generating ML libraries using ASPLs because 1) this task is complicated even for experienced human programmers and 2) there are limited code examples because of the esoteric and evolving nature of ASPLs. Therefore, LLMs need complex reasoning with limited data in order to complete this task. To address these challenges, we introduce an adaptive self-improvement agentic system. In order to evaluate the effectiveness of our system, we construct a benchmark of a typical ML library and generate ASPL code with both open and closed-source LLMs on this benchmark. Our results show improvements of up to 3.9times over a baseline single LLM.

The performance and usability of Large-Language Models (LLMs) are driving their use in explanation generation tasks. However, despite their widespread adoption, LLM explanations have been found to be unreliable, making it difficult for users to distinguish good from bad explanations. To address this issue, we present Rubrik's CUBE, an education-inspired rubric and a dataset of 26k explanations, written and later quality-annotated using the rubric by both humans and six open- and closed-source LLMs. The CUBE dataset focuses on two reasoning and two language tasks, providing the necessary diversity for us to effectively test our proposed rubric. Using Rubrik, we find that explanations are influenced by both task and perceived difficulty. Low quality stems primarily from a lack of conciseness in LLM-generated explanations, rather than cohesion and word choice. The full dataset, rubric, and code will be made available upon acceptance.

Large language models (LLMs) demonstrate strong potential as agents for tool invocation due to their advanced comprehension and planning capabilities. Users increasingly rely on LLM-based agents to solve complex missions through iterative interactions. However, existing benchmarks predominantly access agents in single-mission scenarios, failing to capture real-world complexity. To bridge this gap, we propose the Multi-Mission Tool Bench. In the benchmark, each test case comprises multiple interrelated missions. This design requires agents to dynamically adapt to evolving demands. Moreover, the proposed benchmark explores all possible mission-switching patterns within a fixed mission number. Specifically, we propose a multi-agent data generation framework to construct the benchmark. We also propose a novel method to evaluate the accuracy and efficiency of agent decisions with dynamic decision trees. Experiments on diverse open-source and closed-source LLMs reveal critical factors influencing agent robustness and provide actionable insights to the tool invocation society.

Large Language Models (LLMs) are now increasingly widely used to simulate personas in virtual environments, leveraging their instruction-following capability. However, we discovered that even state-of-the-art LLMs cannot simulate personas with reversed performance (e.g., student personas with low proficiency in educational settings), which impairs the simulation diversity and limits the practical applications of the simulated environments. In this work, using mathematical reasoning as a representative scenario, we propose the first benchmark dataset for evaluating LLMs on simulating personas with reversed performance, a capability that we dub "counterfactual instruction following". We evaluate both open-weight and closed-source LLMs on this task and find that LLMs, including the OpenAI o1 reasoning model, all struggle to follow counterfactual instructions for simulating reversedly performing personas. Intersectionally simulating both the performance level and the race population of a persona worsens the effect even further. These results highlight the challenges of counterfactual instruction following and the need for further research.

Large language models (LLMs) have been widely applied to assist test generation with the source code under test provided as the context. This paper aims to answer the question: If the source code under test is incorrect, will LLMs be misguided when generating tests? The effectiveness of test cases is measured by their accuracy, coverage, and bug detection effectiveness. Our evaluation results with five open- and six closed-source LLMs on four datasets demonstrate that incorrect code can significantly mislead LLMs in generating correct, high-coverage, and bug-revealing tests. For instance, in the HumanEval dataset, LLMs achieve 80.45% test accuracy when provided with task descriptions and correct code, but only 57.12% when given task descriptions and incorrect code. For the APPS dataset, prompts with correct code yield tests that detect 39.85% of the bugs, while prompts with incorrect code detect only 19.61%. These findings have important implications for the deployment of LLM-based testing: using it on mature code may help protect against future regression, but on early-stage immature code, it may simply bake in errors. Our findings also underscore the need for further research to improve LLMs resilience against incorrect code in generating reliable and bug-revealing tests.

The safety alignment of Large Language Models (LLMs) is vulnerable to both manual and automated jailbreak attacks, which adversarially trigger LLMs to output harmful content. However, current methods for jailbreaking LLMs, which nest entire harmful prompts, are not effective at concealing malicious intent and can be easily identified and rejected by well-aligned LLMs. This paper discovers that decomposing a malicious prompt into separated sub-prompts can effectively obscure its underlying malicious intent by presenting it in a fragmented, less detectable form, thereby addressing these limitations. We introduce an automatic prompt Decomposition and Reconstruction framework for jailbreak Attack (DrAttack). DrAttack includes three key components: (a) `Decomposition' of the original prompt into sub-prompts, (b) `Reconstruction' of these sub-prompts implicitly by in-context learning with semantically similar but harmless reassembling demo, and (c) a `Synonym Search' of sub-prompts, aiming to find sub-prompts' synonyms that maintain the original intent while jailbreaking LLMs. An extensive empirical study across multiple open-source and closed-source LLMs demonstrates that, with a significantly reduced number of queries, DrAttack obtains a substantial gain of success rate over prior SOTA prompt-only attackers. Notably, the success rate of 78.0\% on GPT-4 with merely 15 queries surpassed previous art by 33.1\%. The project is available at https://github.com/xirui-li/DrAttack.

Large language models (LLMs) have achieved impressive performance on code generation. Although prior studies enhanced LLMs with prompting techniques and code refinement, they still struggle with complex programming problems due to rigid solution plans. In this paper, we draw on pair programming practices to propose PairCoder, a novel LLM-based framework for code generation. PairCoder incorporates two collaborative LLM agents, namely a Navigator agent for high-level planning and a Driver agent for specific implementation. The Navigator is responsible for proposing promising solution plans, selecting the current optimal plan, and directing the next iteration round based on execution feedback. The Driver follows the guidance of Navigator to undertake initial code generation, code testing, and refinement. This interleaved and iterative workflow involves multi-plan exploration and feedback-based refinement, which mimics the collaboration of pair programmers. We evaluate PairCoder with both open-source and closed-source LLMs on various code generation benchmarks. Extensive experimental results demonstrate the superior accuracy of PairCoder, achieving relative pass@1 improvements of 12.00%-162.43% compared to prompting LLMs directly.

Existing large language models (LLMs) evaluation methods typically focus on testing the performance on some closed-environment and domain-specific benchmarks with human annotations. In this paper, we explore a novel unsupervised evaluation direction, utilizing peer-review mechanisms to measure LLMs automatically. In this setting, both open-source and closed-source LLMs lie in the same environment, capable of answering unlabeled questions and evaluating each other, where each LLM's response score is jointly determined by other anonymous ones. To obtain the ability hierarchy among these models, we assign each LLM a learnable capability parameter to adjust the final ranking. We formalize it as a constrained optimization problem, intending to maximize the consistency of each LLM's capabilities and scores. The key assumption behind is that high-level LLM can evaluate others' answers more accurately than low-level ones, while higher-level LLM can also achieve higher response scores. Moreover, we propose three metrics called PEN, CIN, and LIS to evaluate the gap in aligning human rankings. We perform experiments on multiple datasets with these metrics, validating the effectiveness of the proposed approach.

In the evolving landscape of large language models (LLMs) tailored for software engineering, the need for benchmarks that accurately reflect real-world development scenarios is paramount. Current benchmarks are either too simplistic or fail to capture the multi-tasking nature of software development. To address this, we introduce CoderUJB, a new benchmark designed to evaluate LLMs across diverse Java programming tasks that are executable and reflective of actual development scenarios, acknowledging Java's prevalence in real-world software production. CoderUJB comprises 2,239 programming questions derived from 17 real open-source Java projects and spans five practical programming tasks. Our empirical study on this benchmark investigates the coding abilities of various open-source and closed-source LLMs, examining the effects of continued pre-training in specific programming languages code and instruction fine-tuning on their performance. The findings indicate that while LLMs exhibit strong potential, challenges remain, particularly in non-functional code generation (e.g., test generation and defect detection). Importantly, our results advise caution in the specific programming languages continued pre-training and instruction fine-tuning, as these techniques could hinder model performance on certain tasks, suggesting the need for more nuanced strategies. CoderUJB thus marks a significant step towards more realistic evaluations of programming capabilities in LLMs, and our study provides valuable insights for the future development of these models in software engineering.

Large Language Models (LLMs) and Vision Language Models (VLMs) possess extensive knowledge and exhibit promising reasoning abilities; however, they still struggle to perform well in complex, dynamic environments. Real-world tasks require handling intricate interactions, advanced spatial reasoning, long-term planning, and continuous exploration of new strategies-areas in which we lack effective methodologies for comprehensively evaluating these capabilities. To address this gap, we introduce BALROG, a novel benchmark designed to assess the agentic capabilities of LLMs and VLMs through a diverse set of challenging games. Our benchmark incorporates a range of existing reinforcement learning environments with varying levels of difficulty, including tasks that are solvable by non-expert humans in seconds to extremely challenging ones that may take years to master (e.g., the NetHack Learning Environment). We devise fine-grained metrics to measure performance and conduct an extensive evaluation of several popular open-source and closed-source LLMs and VLMs. Our findings indicate that while current models achieve partial success in the easier games, they struggle significantly with more challenging tasks. Notably, we observe severe deficiencies in vision-based decision-making, as models perform worse when visual representations of the environments are provided. We release BALROG as an open and user-friendly benchmark to facilitate future research and development in the agentic community.

Jailbreaking is an emerging adversarial attack that bypasses the safety alignment deployed in off-the-shelf large language models (LLMs) and has evolved into multiple categories: human-based, optimization-based, generation-based, and the recent indirect and multilingual jailbreaks. However, delivering a practical jailbreak defense is challenging because it needs to not only handle all the above jailbreak attacks but also incur negligible delays to user prompts, as well as be compatible with both open-source and closed-source LLMs. Inspired by how the traditional security concept of shadow stacks defends against memory overflow attacks, this paper introduces a generic LLM jailbreak defense framework called SelfDefend, which establishes a shadow LLM as a defense instance (in detection state) to concurrently protect the target LLM instance (in normal answering state) in the normal stack and collaborate with it for checkpoint-based access control. The effectiveness of SelfDefend builds upon our observation that existing LLMs can identify harmful prompts or intentions in user queries, which we empirically validate using mainstream GPT-3.5/4 models against major jailbreak attacks. To further improve the defense's robustness and minimize costs, we employ a data distillation approach to tune dedicated open-source defense models. When deployed to protect GPT-3.5/4, Claude, Llama-2-7b/13b, and Mistral, these models outperform seven state-of-the-art defenses and match the performance of GPT-4-based SelfDefend, with significantly lower extra delays. Further experiments show that the tuned models are robust to adaptive jailbreaks and prompt injections.

Investigating hallucination issues in large language models (LLMs) within cross-lingual and cross-modal scenarios can greatly advance the large-scale deployment in real-world applications. Nevertheless, the current studies are limited to a single scenario, either cross-lingual or cross-modal, leaving a gap in the exploration of hallucinations in the joint cross-lingual and cross-modal scenarios. Motivated by this, we introduce a novel joint Cross-lingual and Cross-modal Hallucinations benchmark (CCHall) to fill this gap. Specifically, CCHall simultaneously incorporates both cross-lingual and cross-modal hallucination scenarios, which can be used to assess the cross-lingual and cross-modal capabilities of LLMs. Furthermore, we conduct a comprehensive evaluation on CCHall, exploring both mainstream open-source and closed-source LLMs. The experimental results highlight that current LLMs still struggle with CCHall. We hope CCHall can serve as a valuable resource to assess LLMs in joint cross-lingual and cross-modal scenarios.

Benefiting from diverse instruction datasets, contemporary Large Language Models (LLMs) perform effectively as AI assistants in collaborating with humans. However, LLMs still struggle to generate natural and colloquial responses in real-world applications such as chatbots and psychological counseling that require more human-like interactions. To address these limitations, we introduce NICO, a Natural Interactive COnversation dataset in Chinese. We first use GPT-4-turbo to generate dialogue drafts and make them cover 20 daily-life topics and 5 types of social interactions. Then, we hire workers to revise these dialogues to ensure that they are free of grammatical errors and unnatural utterances. We define two dialogue-level natural conversation tasks and two sentence-level tasks for identifying and rewriting unnatural sentences. Multiple open-source and closed-source LLMs are tested and analyzed in detail. The experimental results highlight the challenge of the tasks and demonstrate how NICO can help foster the natural dialogue capabilities of LLMs. The dataset will be released.

Online reviews play a pivotal role in influencing consumer decisions across various domains, from purchasing products to selecting hotels or restaurants. However, the sheer volume of reviews -- often containing repetitive or irrelevant content -- leads to information overload, making it challenging for users to extract meaningful insights. Traditional opinion summarization models face challenges in handling long inputs and large volumes of reviews, while newer Large Language Model (LLM) approaches often fail to generate accurate and faithful summaries. To address those challenges, this paper introduces (1) a new dataset of long-form user reviews, each entity comprising over a thousand reviews, (2) two training-free LLM-based summarization approaches that scale to long inputs, and (3) automatic evaluation metrics. Our dataset of user reviews is paired with in-depth and unbiased critical summaries by domain experts, serving as a reference for evaluation. Additionally, our novel reference-free evaluation metrics provide a more granular, context-sensitive assessment of summary faithfulness. We benchmark several open-source and closed-source LLMs using our methods. Our evaluation reveals that LLMs still face challenges in balancing sentiment and format adherence in long-form summaries, though open-source models can narrow the gap when relevant information is retrieved in a focused manner.

We present Plancraft, a multi-modal evaluation dataset for LLM agents. Plancraft has both a text-only and multi-modal interface, based on the Minecraft crafting GUI. We include the Minecraft Wiki to evaluate tool use and Retrieval Augmented Generation (RAG), as well as an oracle planner and oracle RAG information extractor, to ablate the different components of a modern agent architecture. To evaluate decision-making, Plancraft also includes a subset of examples that are intentionally unsolvable, providing a realistic challenge that requires the agent not only to complete tasks but also to decide whether they are solvable at all. We benchmark both open-source and closed-source LLMs and strategies on our task and compare their performance to a handcrafted planner. We find that LLMs and VLMs struggle with the planning problems that Plancraft introduces, and we offer suggestions on how to improve their capabilities.

In this paper, we introduce MATA, a novel evaluation dataset to assess the ability of Large Language Models (LLMs) in Telugu language, comprising 729 carefully curated multiple-choice and open-ended questions that span diverse linguistic dimensions. We evaluate 11 open-weight and closed-source LLMs on our dataset and present a fine-grained analysis of their performance. Further, we empirically show how LLMs rely on superficial heuristics such as answer position and distractor patterns for multiple-choice questions. Finally, we also compare LLM-as-a-judge evaluation with human evaluation for open-ended questions and draw some conclusions on its reliability in a low-resource language. We argue that such fine-grained evaluation is essential for understanding model limitations and can inform the development of more linguistically capable LLMs, while also serving as a foundation for future research in Telugu NLP.

Critique ability are crucial in the scalable oversight and self-improvement of Large Language Models (LLMs). While many recent studies explore the critique ability of LLMs to judge and refine flaws in generations, how to comprehensively and reliably measure the critique abilities of LLMs is under-explored. This paper introduces \shortname, a novel benchmark designed to comprehensively and reliably evaluate four key critique ability dimensions of LLMs: feedback, comparison, refinement and meta-feedback. \shortname~encompasses nine diverse tasks, each assessing the LLMs' ability to critique responses at varying levels of quality granularity. Our extensive evaluations of open-source and closed-source LLMs reveal intriguing relationships between the critique ability and tasks, response qualities, and model scales. Datasets, resources and evaluation toolkit for \shortname~will be publicly released at https://github.com/gmftbyGMFTBY/CriticBench.

This study investigates the reasoning robustness of large language models (LLMs) on mathematical problem-solving tasks under systematically introduced input perturbations. Using the GSM8K dataset as a controlled testbed, we evaluate how well state-of-the-art models maintain logical consistency and correctness when confronted with four categories of prompt perturbations: irrelevant context, pathological instructions, factually relevant but non-essential context, and a combination of the latter two. Our experiments, conducted on thirteen open-source and closed-source LLMs, reveal that introducing irrelevant context within the model's context window significantly degrades performance, suggesting that distinguishing essential from extraneous details remains a pressing challenge. Surprisingly, performance regressions are relatively insensitive to the complexity of the reasoning task, as measured by the number of steps required, and are not strictly correlated with model size. Moreover, we observe that certain perturbations inadvertently trigger chain-of-thought-like reasoning behaviors, even without explicit prompting. Our findings highlight critical vulnerabilities in current LLMs and underscore the need for improved robustness against noisy, misleading, and contextually dense inputs, paving the way for more resilient and reliable reasoning in real-world applications.

Proteins, as essential biomolecules, play a central role in biological processes, including metabolic reactions and DNA replication. Accurate prediction of their properties and functions is crucial in biological applications. Recent development of protein language models (pLMs) with supervised fine tuning provides a promising solution to this problem. However, the fine-tuned model is tailored for particular downstream prediction task, and achieving general-purpose protein understanding remains a challenge. In this paper, we introduce Structure-Enhanced Protein Instruction Tuning (SEPIT) framework to bridge this gap. Our approach integrates a noval structure-aware module into pLMs to inform them with structural knowledge, and then connects these enhanced pLMs to large language models (LLMs) to generate understanding of proteins. In this framework, we propose a novel two-stage instruction tuning pipeline that first establishes a basic understanding of proteins through caption-based instructions and then refines this understanding using a mixture of experts (MoEs) to learn more complex properties and functional information with the same amount of activated parameters. Moreover, we construct the largest and most comprehensive protein instruction dataset to date, which allows us to train and evaluate the general-purpose protein understanding model. Extensive experimental results on open-ended generation and closed-set answer tasks demonstrate the superior performance of SEPIT over both closed-source general LLMs and open-source LLMs trained with protein knowledge.

Multi-turn instruction following capability constitutes a core competency of large language models (LLMs) in real-world applications. Existing evaluation benchmarks predominantly focus on fine-grained constraint satisfaction and domain-specific capability assessment, yet overlook the crucial structural dependency between dialogue turns that distinguishes multi-turn from single-turn interactions. This structural dependency not only reflects user intent but also establishes a second dimension for instruction following evaluation beyond constraint satisfaction. To address this gap, we propose StructFlowBench, a multi-turn instruction following benchmark with structural flow modeling. The benchmark innovatively defines a structural flow framework comprising six fundamental inter-turn relationships, which not only introduces novel structural constraints for model evaluation but also serves as generation parameters for creating customized dialogue flows tailored to specific scenarios. Adopting established LLM-based automatic evaluation methodologies, we conduct systematic evaluations of 13 leading open-source and closed-source LLMs. Experimental results reveal significant deficiencies in current models' comprehension of multi-turn dialogue structures. The code is available at https://github.com/MLGroupJLU/StructFlowBench.

Despite the notable advancements of existing prompting methods, such as In-Context Learning and Chain-of-Thought for Large Language Models (LLMs), they still face challenges related to various biases. Traditional debiasing methods primarily focus on the model training stage, including approaches based on data augmentation and reweighting, yet they struggle with the complex biases inherent in LLMs. To address such limitations, the causal relationship behind the prompting methods is uncovered using a structural causal model, and a novel causal prompting method based on front-door adjustment is proposed to effectively mitigate LLMs biases. In specific, causal intervention is achieved by designing the prompts without accessing the parameters and logits of LLMs. The chain-of-thought generated by LLM is employed as the mediator variable and the causal effect between input prompts and output answers is calculated through front-door adjustment to mitigate model biases. Moreover, to accurately represent the chain-of-thoughts and estimate the causal effects, contrastive learning is used to fine-tune the encoder of chain-of-thought by aligning its space with that of the LLM. Experimental results show that the proposed causal prompting approach achieves excellent performance across seven natural language processing datasets on both open-source and closed-source LLMs.

Large language models (LLMs) have garnered significant attention and widespread usage due to their impressive performance in various tasks. However, they are not without their own set of challenges, including issues such as hallucinations, factual inconsistencies, and limitations in numerical-quantitative reasoning. Evaluating LLMs in miscellaneous reasoning tasks remains an active area of research. Prior to the breakthrough of LLMs, Transformers had already proven successful in the medical domain, effectively employed for various natural language understanding (NLU) tasks. Following this trend, LLMs have also been trained and utilized in the medical domain, raising concerns regarding factual accuracy, adherence to safety protocols, and inherent limitations. In this paper, we focus on evaluating the natural language inference capabilities of popular open-source and closed-source LLMs using clinical trial reports as the dataset. We present the performance results of each LLM and further analyze their performance on a development set, particularly focusing on challenging instances that involve medical abbreviations and require numerical-quantitative reasoning. Gemini, our leading LLM, achieved a test set F1-score of 0.748, securing the ninth position on the task scoreboard. Our work is the first of its kind, offering a thorough examination of the inference capabilities of LLMs within the medical domain.

Prompt injection poses serious security risks to real-world LLM applications, particularly autonomous agents. Although many defenses have been proposed, their robustness against adaptive attacks remains insufficiently evaluated, potentially creating a false sense of security. In this work, we propose PISmith, a reinforcement learning (RL)-based red-teaming framework that systematically assesses existing prompt-injection defenses by training an attack LLM to optimize injected prompts in a practical black-box setting, where the attacker can only query the defended LLM and observe its outputs. We find that directly applying standard GRPO to attack strong defenses leads to sub-optimal performance due to extreme reward sparsity -- most generated injected prompts are blocked by the defense, causing the policy's entropy to collapse before discovering effective attack strategies, while the rare successes cannot be learned effectively. In response, we introduce adaptive entropy regularization and dynamic advantage weighting to sustain exploration and amplify learning from scarce successes. Extensive evaluation on 13 benchmarks demonstrates that state-of-the-art prompt injection defenses remain vulnerable to adaptive attacks. We also compare PISmith with 7 baselines across static, search-based, and RL-based attack categories, showing that PISmith consistently achieves the highest attack success rates. Furthermore, PISmith achieves strong performance in agentic settings on InjecAgent and AgentDojo against both open-source and closed-source LLMs (e.g., GPT-4o-mini and GPT-5-nano). Our code is available at https://github.com/albert-y1n/PISmith.

Large language models (LLMs) have succeeded significantly in various applications but remain susceptible to adversarial jailbreaks that void their safety guardrails. Previous attempts to exploit these vulnerabilities often rely on high-cost computational extrapolations, which may not be practical or efficient. In this paper, inspired by the authority influence demonstrated in the Milgram experiment, we present a lightweight method to take advantage of the LLMs' personification capabilities to construct a virtual, nested scene, allowing it to realize an adaptive way to escape the usage control in a normal scenario. Empirically, the contents induced by our approach can achieve leading harmfulness rates with previous counterparts and realize a continuous jailbreak in subsequent interactions, which reveals the critical weakness of self-losing on both open-source and closed-source LLMs, e.g., Llama-2, Llama-3, GPT-3.5, GPT-4, and GPT-4o. The code and data are available at: https://github.com/tmlr-group/DeepInception.

Recent advancements in Large Language Models (LLMs) have shown significant progress in understanding complex natural language. One important application of LLM is LLM-based AI Agent, which leverages the ability of LLM as well as external tools for complex-task solving. To make sure LLM Agents follow an effective and reliable procedure to solve the given task, manually designed workflows are usually used to guide the working mechanism of agents. However, manually designing the workflows requires considerable efforts and domain knowledge, making it difficult to develop and deploy agents on massive scales. To address these issues, we propose AutoFlow, a framework designed to automatically generate workflows for agents to solve complex tasks. AutoFlow takes natural language program as the format of agent workflow and employs a workflow optimization procedure to iteratively optimize the workflow quality. Besides, this work offers two workflow generation methods: fine-tuning-based and in-context-based methods, making the AutoFlow framework applicable to both open-source and closed-source LLMs. Experimental results show that our framework can produce robust and reliable agent workflows. We believe that the automatic generation and interpretation of workflows in natural language represent a promising paradigm for solving complex tasks, particularly with the rapid development of LLMs. The source code of this work is available at https://github.com/agiresearch/AutoFlow.

Large language models (LLMs) achieve strong performance across a wide range of tasks, but remain frozen after pretraining until subsequent updates. Many real-world applications require timely, domain-specific information, motivating the need for efficient mechanisms to incorporate new knowledge. In this paper, we introduce MeMo (Memory as a Model), a modular framework that encodes new knowledge into a dedicated memory model while keeping the LLM parameters unchanged. Compared to existing methods, MeMo offers several advantages: (a) it captures complex cross-document relationships, (b) it is robust to retrieval noise, (c) it avoids catastrophic forgetting in the LLM, (d) it does not require access to the LLM's weights or output logits, enabling plug-and-play integration with both open and proprietary closed-source LLMs, and (e) its retrieval cost is independent of corpus size at inference time. Our experimental results on three benchmarks, BrowseComp-Plus, NarrativeQA, and MuSiQue, show that MeMo achieves strong performance compared to existing methods across diverse settings.

Iterative industrial design-simulation optimization is bottlenecked by the CAD-CAE semantic gap: translating simulation feedback into valid geometric edits under diverse, coupled constraints. To fill this gap, we propose COSMO-Agent (Closed-loop Optimization, Simulation, and Modeling Orchestration), a tool-augmented reinforcement learning (RL) framework that teaches LLMs to complete the closed-loop CAD-CAE process. Specifically, we cast CAD generation, CAE solving, result parsing, and geometry revision as an interactive RL environment, where an LLM learns to orchestrate external tools and revise parametric geometries until constraints are satisfied. To make this learning stable and industrially usable, we design a multi-constraint reward that jointly encourages feasibility, toolchain robustness, and structured output validity. In addition, we contribute an industry-aligned dataset that covers 25 component categories with executable CAD-CAE tasks to support realistic training and evaluation. Experiments show that COSMO-Agent training substantially improves small open-source LLMs for constraint-driven design, exceeding large open-source and strong closed-source models in feasibility, efficiency, and stability.

Planning is a fundamental capability for large language models (LLMs) because such complex tasks require models to coordinate goals, constraints, resources, and long-term consequences into executable and verifiable solutions. Existing planning benchmarks, however, usually treat planning data as fixed collections of instances rather than controllable generation targets. This limits scenario coverage, ties difficulty to surface-level proxies rather than structural sources, and offers limited support for scalable generation, automatic verification, or planning-oriented training. We introduce PlanningBench, a framework for generating scalable, diverse, and verifiable planning data for both evaluation and training. PlanningBench starts from real planning scenarios and abstracts practical workflows into a structured taxonomy of more than 30 task types, subtasks, constraint families, and difficulty factors. Guided by this taxonomy, a constraint-driven synthesis pipeline instantiates self-contained planning problems with adaptive difficulty control, quality filtering, and instance-level verification checklists. This shifts planning data construction from fixed benchmark collection to controllable generation while preserving realistic task grounding. We use PlanningBench to evaluate open-source and closed-source frontier LLMs, and find that current models still struggle to produce complete solutions under coupled constraints. Beyond evaluation, reinforcement learning on verified PlanningBench data improves performance on unseen planning benchmarks and broader instruction-following tasks. Further analysis suggests that determinate or well-specified optimal solutions provide clearer reward signals and more stable training dynamics. Overall, PlanningBench provides a controllable source of planning data for diagnosing and improving generalizable planning abilities in LLMs.

The past year has seen rapid acceleration in the development of large language models (LLMs). However, without proper steering and safeguards, LLMs will readily follow malicious instructions, provide unsafe advice, and generate toxic content. We introduce SimpleSafetyTests (SST) as a new test suite for rapidly and systematically identifying such critical safety risks. The test suite comprises 100 test prompts across five harm areas that LLMs, for the vast majority of applications, should refuse to comply with. We test 11 open-access and open-source LLMs and four closed-source LLMs, and find critical safety weaknesses. While some of the models do not give a single unsafe response, most give unsafe responses to more than 20% of the prompts, with over 50% unsafe responses in the extreme. Prepending a safety-emphasising system prompt substantially reduces the occurrence of unsafe responses, but does not completely stop them from happening. Trained annotators labelled every model response to SST (n = 3,000). We use these annotations to evaluate five AI safety filters (which assess whether a models' response is unsafe given a prompt) as a way of automatically evaluating models' performance on SST. The filters' performance varies considerably. There are also differences across the five harm areas, and on the unsafe versus safe responses. The widely-used Perspective API has 72% accuracy and a newly-created zero-shot prompt to OpenAI's GPT-4 performs best with 89% accuracy. Content Warning: This paper contains prompts and responses that relate to child abuse, suicide, self-harm and eating disorders, scams and fraud, illegal items, and physical harm.

While large language models (LLMs) have exhibited impressive conversational capabilities, their proficiency in delivering personalized responses remains unclear. Although recent benchmarks automatically evaluate persona consistency in role-playing contexts using LLM-based judgment, the evaluation of personalization in response generation remains underexplored. To address this gap, we present an automated benchmarking pipeline, PersoBench, to evaluate the personalization ability of LLMs in persona-aware dialogue generation within a zero-shot setting. Our framework employs a structured pipeline comprising speaker-aware annotation, task-specific and context-driven prompt construction, response post-processing, and automated evaluation across multiple dimensions of generation quality. In particular, the pipeline performs text preprocessing and speaker labeling, constructs structured prompts with task instructions and LLM roles, validates response format, and evaluates valid outputs across fluency, personalization, diversity, and coherence. We assess the performance of four open-source and four closed-source LLMs using well-known datasets and a range of explicit metrics. Our findings reveal that while LLMs excel at generating fluent and diverse responses, they are far from satisfactory in delivering personalized and coherent responses, considering both the conversation context and the provided personas.

Chat template is a common technique used in the training and inference stages of Large Language Models (LLMs). It can transform input and output data into role-based and templated expressions to enhance the performance of LLMs. However, this also creates a breeding ground for novel attack surfaces. In this paper, we first reveal that the customizability of chat templates allows an attacker who controls the template to inject arbitrary strings into the system prompt without the user's notice. Building on this, we propose a training-free backdoor attack, termed BadTemplate. Specifically, BadTemplate inserts carefully crafted malicious instructions into the high-priority system prompt, thereby causing the target LLM to exhibit persistent backdoor behaviors. BadTemplate outperforms traditional backdoor attacks by embedding malicious instructions directly into the system prompt, eliminating the need for model retraining while achieving high attack effectiveness with minimal cost. Furthermore, its simplicity and scalability make it easily and widely deployed in real-world systems, raising serious risks of rapid propagation, economic damage, and large-scale misinformation. Furthermore, detection by major third-party platforms HuggingFace and LLM-as-a-judge proves largely ineffective against BadTemplate. Extensive experiments conducted on 5 benchmark datasets across 6 open-source and 3 closed-source LLMs, compared with 3 baselines, demonstrate that BadTemplate achieves up to a 100% attack success rate and significantly outperforms traditional prompt-based backdoors in both word-level and sentence-level attacks. Our work highlights the potential security risks raised by chat templates in the LLM supply chain, thereby supporting the development of effective defense mechanisms.

Code large language models (LLMs) have made significant progress in code debugging by directly generating the correct code based on the buggy code snippet. Programming benchmarks, typically consisting of buggy code snippet and their associated test cases, are used to assess the debugging capabilities of LLMs. However, many existing benchmarks primarily focus on Python and are often limited in terms of language diversity (e.g., DebugBench and DebugEval). To advance the field of multilingual debugging with LLMs, we propose the first massively multilingual debugging benchmark, which includes 3.6K test samples of 18 programming languages and covers the automated program repair (APR) task, the code review (CR) task, and the bug identification (BI) task. Further, we introduce the debugging instruction corpora MDEVAL-INSTRUCT by injecting bugs into the correct multilingual queries and solutions (xDebugGen). Further, a multilingual debugger xDebugCoder trained on MDEVAL-INSTRUCT as a strong baseline specifically to handle the bugs of a wide range of programming languages (e.g. "Missing Mut" in language Rust and "Misused Macro Definition" in language C). Our extensive experiments on MDEVAL reveal a notable performance gap between open-source models and closed-source LLMs (e.g., GPT and Claude series), highlighting huge room for improvement in multilingual code debugging scenarios.

Code large language models (LLMs) have shown remarkable advances in code understanding, completion, and generation tasks. Programming benchmarks, comprised of a selection of code challenges and corresponding test cases, serve as a standard to evaluate the capability of different LLMs in such tasks. However, most existing benchmarks primarily focus on Python and are still restricted to a limited number of languages, where other languages are translated from the Python samples (e.g. MultiPL-E) degrading the data diversity. To further facilitate the research of code LLMs, we propose a massively multilingual code benchmark covering 40 programming languages (McEval) with 16K test samples, which substantially pushes the limits of code LLMs in multilingual scenarios. The benchmark contains challenging code completion, understanding, and generation evaluation tasks with finely curated massively multilingual instruction corpora McEval-Instruct. In addition, we introduce an effective multilingual coder mCoder trained on McEval-Instruct to support multilingual programming language generation. Extensive experimental results on McEval show that there is still a difficult journey between open-source models and closed-source LLMs (e.g. GPT-series models) in numerous languages. The instruction corpora, evaluation benchmark, and leaderboard are available at https://mceval.github.io/.

While large language models (LLMs) have exhibited impressive instruction-following capabilities, it is still unclear whether and to what extent they can respond to explicit constraints that might be entailed in various instructions. As a significant aspect of LLM alignment, it is thus important to formulate such a specialized set of instructions as well as investigate the resulting behavior of LLMs. To address this vacancy, we propose a new benchmark CoDI-Eval to systematically and comprehensively evaluate LLMs' responses to instructions with various constraints. We construct a large collection of constraints-attributed instructions as a test suite focused on both generalization and coverage. Specifically, we advocate an instruction diversification process to synthesize diverse forms of constraint expression and also deliberate the candidate task taxonomy with even finer-grained sub-categories. Finally, we automate the entire evaluation process to facilitate further developments. Different from existing studies on controllable text generation, CoDI-Eval extends the scope to the prevalent instruction-following paradigm for the first time. We provide extensive evaluations of representative LLMs (e.g., ChatGPT, Vicuna) on CoDI-Eval, revealing their limitations in following instructions with specific constraints and there is still a significant gap between open-source and commercial closed-source LLMs. We believe this benchmark will facilitate research into improving the controllability of LLMs' responses to instructions. Our data and code are available at https://github.com/Xt-cyh/CoDI-Eval.

Large Language Models (LLMs) are distinguished by their architecture, which dictates their parameter size and performance capabilities. Social scientists have increasingly adopted LLMs for text classification tasks, which are difficult to scale with human coders. While very large, closed-source models often deliver superior performance, their use presents significant risks. These include lack of transparency, potential exposure of sensitive data, challenges to replicability, and dependence on proprietary systems. Additionally, their high costs make them impractical for large-scale research projects. In contrast, open-source models, although available in various sizes, may underperform compared to commercial alternatives if used without further fine-tuning. However, open-source models offer distinct advantages: they can be run locally (ensuring data privacy), fine-tuned for specific tasks, shared within the research community, and integrated into reproducible workflows. This study demonstrates that small, fine-tuned open-source LLMs can achieve equal or superior performance to models such as ChatGPT-4. We further explore the relationship between training set size and fine-tuning efficacy in open-source models. Finally, we propose a hybrid workflow that leverages the strengths of both open and closed models, offering a balanced approach to performance, transparency, and reproducibility.

Many Bangla words are at once personal names and culturally loaded common nouns, "Maya" is both a girl's name and a word for affectionate compassion. Choosing the right reading demands cultural knowledge that is scarce in the pretraining data of modern language models. We introduce Culturally Entangled Homograph (CEH) disambiguation and build a Bangla benchmark of 1,516 expert-verified sentences (3,032 labelled occurrences) in which one word appears twice with two distinct readings, each labelled with a culturally grounded category and an explanation of the reasoning behind it. Across open- and closed-source models, we find a systematic dominant-meaning bias: models default to the common-noun sense and overlook the name. A Bangla-specific model fails under every prompting regime we test, showing that language-specific pretraining alone does not confer cultural grounding. We further show that contrastive chain-of-thought prompting can sharply reduce this bias without training, and that distilling cultural explanations teaches small (1-3B) models to reason toward the correct reading rather than memorise labels, cutting dominant-meaning bias from as high as 100% to under 5% and turning the failed Bangla-specific model into our strongest system. Dataset and code are available at https://github.com/ashuvo25/BanglaCEH.

Real-world conversations are rich with pragmatic elements, such as entity mentions, references, and implicatures. Understanding such nuances is a requirement for successful natural communication, and often requires building a local world model which encodes such elements and captures the dynamics of their evolving states. However, it is not well-understood whether language models (LMs) construct or maintain a robust implicit representation of conversations. In this work, we evaluate the ability of LMs to encode and update their internal world model in dyadic conversations and test their malleability under linguistic alterations. To facilitate this, we apply seven minimal linguistic alterations to conversations sourced from popular datasets and construct two benchmarks comprising yes-no questions. We evaluate a wide range of open and closed source LMs and observe that they struggle to maintain robust accuracy. Our analysis unveils that LMs struggle to memorize crucial details, such as tracking entities under linguistic alterations to conversations. We then propose a dual-perspective interpretability framework which identifies transformer layers that are useful or harmful and highlights linguistic alterations most influenced by harmful layers, typically due to encoding spurious signals or relying on shortcuts. Inspired by these insights, we propose two layer-regularization based fine-tuning strategies that suppress the effect of the harmful layers.

Large language models (LLMs) have exhibited their problem-solving abilities in mathematical reasoning. Solving realistic optimization (OPT) problems in application scenarios requires advanced and applied mathematics ability. However, current OPT benchmarks that merely solve linear programming are far from complex realistic situations. In this work, we propose OptiBench, a benchmark for End-to-end optimization problem-solving with human-readable inputs and outputs. OptiBench contains rich optimization problems, including linear and nonlinear programming with or without tabular data, which can comprehensively evaluate LLMs' solving ability. In our benchmark, LLMs are required to call a code solver to provide precise numerical answers. Furthermore, to alleviate the data scarcity for optimization problems, and to bridge the gap between open-source LLMs on a small scale (e.g., Llama-3-8b) and closed-source LLMs (e.g., GPT-4), we further propose a data synthesis method namely ReSocratic. Unlike general data synthesis methods that proceed from questions to answers, \ReSocratic first incrementally synthesizes formatted optimization demonstration with mathematical formulations step by step and then back-translates the generated demonstrations into questions. Based on this, we synthesize the ReSocratic-29k dataset. We further conduct supervised fine-tuning with ReSocratic-29k on multiple open-source models. Experimental results show that ReSocratic-29k significantly improves the performance of open-source models.

Despite the strong language understanding abilities of large language models (LLMs), they still struggle with reliable question answering (QA) over long, structured documents, particularly for numerical reasoning. Financial annual reports exemplify this difficulty: financial statement analysis often hinges on accurate arithmetic, and analysts derive key indicators by integrating evidence scattered across multiple tables and narrative text. However, existing benchmarks focus largely on single-table settings, leaving cross-table document-level numerical reasoning underexplored. To address this gap, we introduce FinLongDocQA, a dataset for both single-table and cross-table financial numerical reasoning in long-context reports. Evaluating both closed-source and open-source LLMs on FinLongDocQA reveals two bottlenecks: (1) annual reports often exceed 129k tokens, exacerbating the context rot problem for locating relevant tables; and (2) even when relevant evidence is located, LLMs remain prone to errors in multi-step numerical reasoning. We propose FinLongDocAgent, a Multi-Agent Multi-Round Retrieval-Augmented Generation (RAG) approach that iteratively retrieves evidence, performs intermediate calculations, and verifies results across rounds. Experiments highlight the importance of iterative retrieval and verification for reliable numerical QA in long financial documents.

Large language models (LLMs) have recently advanced text-driven 3D generation, yet Text-to-CAD remains far from supporting industrial product design. Existing benchmarks focus primarily on generating single-part CAD models and evaluate them using geometric similarity metrics that fail to capture functionality, manufacturability, and assemblability. To address this gap, we introduce MUSE, a Text-to-CAD benchmark focused on complex, editable boundary representation (B-Rep) assemblies. MUSE pairs practical design instances with structured Design Specifications and evaluates generated models through a three-stage protocol: code check, geometric check, and design-intent alignment. The final stage uses design-specific rubrics to assess functionality, manufacturability, and assemblability, moving beyond shape matching toward practical design quality. To enable scalable evaluation, we use a rubric-based visual language model (VLM) judge and validate its reliability through human annotation. Experiments on closed-source and open-source LLMs reveal a clear failure cascade from executable code to valid geometry and finally to engineering-ready design, with even the strongest models achieving limited success on fine-grained engineering criteria. Together, MUSE provides a realistic benchmark and evaluation framework for advancing Text-to-CAD from geometric generation toward true engineering design. Our project website, including the leaderboard, dataset, and code, is available at https://dong7313.github.io/muse-benchmark/.

Multimodal large language models (MLLMs) are increasingly used as OCR systems in document and knowledge-processing pipelines, but their ability to faithfully read real handwriting remains underexplored. Existing OCR benchmarks focus largely on printed text or clean single-line inputs, leaving limited coverage of realistic handwritten OCR scenarios such as multilingual handwriting, writer errors, and structurally complex mathematical expressions. We introduce OmniHandwritingOCR, a diagnostic benchmark for evaluating MLLMs and OCR systems on handwritten OCR. It covers handwritten text recognition and handwritten mathematical expression recognition across six subtasks and twelve subsets, totaling 77.57K labeled images from public datasets and newly collected student writings. A key component is a difficulty-stratified multi-line formula corpus designed to test robustness under increasing structural complexity. We evaluate thirteen open- and closed-source systems with five complementary metrics under a unified protocol. Results show that current systems remain far from faithful transcription: performance drops sharply on complex multi-line formulas, model rankings vary across language and formula settings, and several generative models hallucinate plausible but visually unsupported corrections. OmniHandwritingOCR provides a challenging testbed for diagnosing language, content, structural, and visual-grounding failure modes of multimodal models in handwritten OCR scenarios.

We introduce LLM-Wikirace, a benchmark for evaluating planning, reasoning, and world knowledge in large language models (LLMs). In LLM-Wikirace, models must efficiently navigate Wikipedia hyperlinks step by step to reach a target page from a given source, requiring look-ahead planning and the ability to reason about how concepts are connected in the real world. We evaluate a broad set of open- and closed-source models, including Gemini-3, GPT-5, and Claude Opus 4.5, which achieve the strongest results on the easy level of the task and demonstrate superhuman performance. Despite this, performance drops sharply on hard difficulty: the best-performing model, Gemini-3, succeeds in only 23\% of hard games, highlighting substantial remaining challenges for frontier models. Our analysis shows that world knowledge is a necessary ingredient for success, but only up to a point, beyond this threshold, planning and long-horizon reasoning capabilities become the dominant factors. Trajectory-level analysis further reveals that even the strongest models struggle to replan after failure, frequently entering loops rather than recovering. LLM-Wikirace is a simple benchmark that reveals clear limitations in current reasoning systems, offering an open arena where planning-capable LLMs still have much to prove. Our code and leaderboard available at https:/llmwikirace.github.io.

Code-switching (CS) poses a significant challenge for Large Language Models (LLMs), yet its comprehensibility remains underexplored in LLMs. We introduce CS-Sum, to evaluate the comprehensibility of CS by the LLMs through CS dialogue to English summarization. CS-Sum is the first benchmark for CS dialogue summarization across Mandarin-English (EN-ZH), Tamil-English (EN-TA), and Malay-English (EN-MS), with 900-1300 human-annotated dialogues per language pair. Evaluating ten LLMs, including open and closed-source models, we analyze performance across few-shot, translate-summarize, and fine-tuning (LoRA, QLoRA on synthetic data) approaches. Our findings show that though the scores on automated metrics are high, LLMs make subtle mistakes that alter the complete meaning of the dialogue. To this end, we introduce 3 most common type of errors that LLMs make when handling CS input. Error rates vary across CS pairs and LLMs, with some LLMs showing more frequent errors on certain language pairs, underscoring the need for specialized training on code-switched data.

Adversarial threats against LLMs are escalating faster than current defenses can adapt. We expose a critical geometric blind spot in alignment: adversarial prompts exploit latent camouflage, embedding perilously close to the safe representation manifold while encoding unsafe intent thereby evading surface level defenses like Direct Preference Optimization (DPO), which remain blind to the latent geometry. We introduce ALKALI, the first rigorously curated adversarial benchmark and the most comprehensive to date spanning 9,000 prompts across three macro categories, six subtypes, and fifteen attack families. Evaluation of 21 leading LLMs reveals alarmingly high Attack Success Rates (ASRs) across both open and closed source models, exposing an underlying vulnerability we term latent camouflage, a structural blind spot where adversarial completions mimic the latent geometry of safe ones. To mitigate this vulnerability, we introduce GRACE - Geometric Representation Aware Contrastive Enhancement, an alignment framework coupling preference learning with latent space regularization. GRACE enforces two constraints: latent separation between safe and adversarial completions, and adversarial cohesion among unsafe and jailbreak behaviors. These operate over layerwise pooled embeddings guided by a learned attention profile, reshaping internal geometry without modifying the base model, and achieve up to 39% ASR reduction. Moreover, we introduce AVQI, a geometry aware metric that quantifies latent alignment failure via cluster separation and compactness. AVQI reveals when unsafe completions mimic the geometry of safe ones, offering a principled lens into how models internally encode safety. We make the code publicly available at https://anonymous.4open.science/r/alkali-B416/README.md.

Large language models (LLMs) often exhibit systematic preferences for certain answer choices when responding to multiple-choice questions-a behavior known as selection bias. This bias reduces the accuracy and reliability of LLM outputs, limiting their usefulness in decision-critical applications. While prior work has focused on adjusting model inputs or outputs to mitigate this issue, our work takes a fundamentally different approach by identifying and removing the internal sources of bias. We introduce two methods: Bias Node Pruning (BNP), which prunes parameters that contribute to selection bias, and Auxiliary Option Injection (AOI), which introduces an additional answer choice to reduce bias in both white-box and black-box settings. To address the shortcomings of existing evaluation metrics, we propose Choice Kullback-Leibler Divergence (CKLD), a new metric that captures distributional imbalances in model predictions. Experiments on three LLMs across multiple datasets demonstrate that our methods consistently improve answer accuracy while reducing selection bias, providing a robust solution for both open- and closed-source models.

The rise of Large Language Models (LLMs) is reshaping multimodel models, with speech synthesis being a prominent application. However, existing approaches often underutilize the linguistic intelligence of these models, typically failing to leverage their powerful instruction-following capabilities. This limitation hinders the model's ability to follow text instructions for controllable Text-to-Speech~(TTS). To address this, we propose a new paradigm inspired by ``operationalism'' that decouples instruction understanding from speech generation. We introduce BatonVoice, a framework where an LLM acts as a ``conductor'', understanding user instructions and generating a textual ``plan'' -- explicit vocal features (e.g., pitch, energy). A separate TTS model, the ``orchestra'', then generates the speech from these features. To realize this component, we develop BatonTTS, a TTS model trained specifically for this task. Our experiments demonstrate that BatonVoice achieves strong performance in controllable and emotional speech synthesis, outperforming strong open- and closed-source baselines. Notably, our approach enables remarkable zero-shot cross-lingual generalization, accurately applying feature control abilities to languages unseen during post-training. This demonstrates that objectifying speech into textual vocal features can more effectively unlock the linguistic intelligence of LLMs.

This research introduces a novel evaluation framework designed to assess large language models' (LLMs) ability to acknowledge uncertainty on 675 fundamentally unsolvable problems. Using a curated dataset of graduate-level grand challenge questions with intentionally unknowable answers, we evaluated twelve state-of-the-art LLMs, including both open and closed-source models, on their propensity to admit ignorance rather than generate plausible but incorrect responses. The best models scored in 62-68% accuracy ranges for admitting the problem solution was unknown in fields ranging from biology to philosophy and mathematics. We observed an inverse relationship between problem difficulty and model accuracy, with GPT-4 demonstrating higher rates of uncertainty acknowledgment on more challenging problems (35.8%) compared to simpler ones (20.0%). This pattern indicates that models may be more prone to generate speculative answers when problems appear more tractable. The study also revealed significant variations across problem categories, with models showing difficulty in acknowledging uncertainty in invention and NP-hard problems while performing relatively better on philosophical and psychological challenges. These results contribute to the growing body of research on artificial general intelligence (AGI) assessment by highlighting the importance of uncertainty recognition as a critical component of future machine intelligence evaluation. This impossibility test thus extends previous theoretical frameworks for universal intelligence testing by providing empirical evidence of current limitations in LLMs' ability to recognize their own knowledge boundaries, suggesting new directions for improving model training architectures and evaluation approaches.

Text-to-SQL models allow users to interact with a database more easily by generating executable SQL statements from natural-language questions. Despite recent successes on simpler databases and questions, current Text-to-SQL methods still suffer from low execution accuracy on industry-scale databases and complex questions involving domain-specific business logic. We present PaVeRL-SQL, a framework that combines Partial-Match Rewards and Verbal Reinforcement Learning to drive self-improvement in reasoning language models (RLMs) for Text-to-SQL. To handle practical use cases, we adopt two pipelines: (1) a newly designed in-context learning framework with group self-evaluation (verbal-RL), using capable open- and closed-source large language models (LLMs) as backbones; and (2) a chain-of-thought (CoT) RL pipeline with a small backbone model (OmniSQL-7B) trained with a specially designed reward function and two-stage RL. These pipelines achieve state-of-the-art (SOTA) results on popular Text-to-SQL benchmarks -- Spider, Spider 2.0, and BIRD. For the industrial-level Spider2.0-SQLite benchmark, the verbal-RL pipeline achieves an execution accuracy 7.4\% higher than SOTA, and the CoT pipeline is 1.4\% higher. RL training with mixed SQL dialects yields strong, threefold gains, particularly for dialects with limited training data. Overall, PaVeRL-SQL delivers reliable, SOTA Text-to-SQL under realistic industrial constraints. The code is available at https://github.com/PaVeRL-SQL/PaVeRL-SQL.

Large Language Models (LLMs) have revolutionized natural language processing but remain vulnerable to jailbreak attacks, especially multi-turn jailbreaks that distribute malicious intent across benign exchanges and bypass alignment mechanisms. Existing approaches often explore the adversarial space poorly, rely on hand-crafted heuristics, or lack systematic query refinement. We present NEXUS (Network Exploration for eXploiting Unsafe Sequences), a modular framework for constructing, refining, and executing optimized multi-turn attacks. NEXUS comprises: (1) ThoughtNet, which hierarchically expands a harmful intent into a structured semantic network of topics, entities, and query chains; (2) a feedback-driven Simulator that iteratively refines and prunes these chains through attacker-victim-judge LLM collaboration using harmfulness and semantic-similarity benchmarks; and (3) a Network Traverser that adaptively navigates the refined query space for real-time attacks. This pipeline uncovers stealthy, high-success adversarial paths across LLMs. On several closed-source and open-source LLMs, NEXUS increases attack success rate by 2.1% to 19.4% over prior methods. Code: https://github.com/inspire-lab/NEXUS

Large language model (LLM) agents have exhibited strong problem-solving competence across domains like research and coding. Yet, it remains underexplored whether LLM agents can tackle compounding real-world problems that require a diverse set of tools to complete. Given a broad, heterogeneous tool repository, LLM agents must not only select appropriate tools based on task planning analysis but also strategically schedule the execution order to ensure efficiency. This paper introduces TPS-Bench to benchmark the ability of LLM agents in solving such problems that demand Tool Planning and Scheduling. TPS-Bench collects 200 compounding tasks of two difficulty levels, based on a tool repository containing hundreds of model context protocol (MCP) tools. In particular, each task is composed of multiple subtasks, such as web search, map navigation, calendar checking, etc., and each subtask can be completed by a basic tool. Our evaluation emphasizes both task completion rate and efficiency. The empirical studies on popular closed-source and open-source LLMs indicate that most models can perform reasonable tool planning, but differ in scheduling. For example, GLM-4.5 achieves an outperforming task completion rate of 64.72% with extensive sequential tool calls, hence suffering from significantly long execution time. By contrast, GPT-4o prioritizes parallel tool calls but achieves only a 45.08% completion rate. Considering reinforcement learning (RL) can be a viable way to improve the scheduling efficiency without compromising performance, we perform an initial study on Qwen3-1.7B and witness a 14% reduction in execution time alongside a 6% gain in task completion rate based on rarely 100 RL training samples. Our code is available https://github.com/hanwenxu1/mcp-agent.

Large Language Models (LLMs) are increasingly deployed in multi-agent settings that require coordination without communication, from human-AI interaction to safety-critical scenarios. Humans often overcome the absence of communication through focal points: salient solutions that naturally stand out to all participants. We present the first large-scale evaluation of how, when, and why focal points emerge in LLMs, comparing their behaviour with humans across cooperative and competitive games, including realistic search and rescue scenarios, demonstrating when focal points enable effective coordination. Across more than 20 open- and closed-source models, we find that LLMs exhibit a remarkable ability to coordinate without communication, often matching or outperforming humans. However, the same models consistently fail in tasks requiring numerical common sense or culturally nuanced notions of salience. We additionally evaluate simple learning-free strategies that substantially improve coordination both among LLMs and between humans and LLMs. Our results reveal striking coordination capabilities, as well as social limitations in modern LLMs, and offer new insight into the latent notions of salience encoded within them. Our findings caution against assuming that LLMs share humans' cultural and perceptual substrate when deployed in coordination settings.

Recent advancements in instruction fine-tuning, alignment methods such as reinforcement learning from human feedback (RLHF), and optimization techniques like direct preference optimization (DPO) have significantly enhanced the adaptability of large language models (LLMs) to user preferences. However, despite these innovations, many LLMs continue to exhibit biases toward Western, Anglo-centric, or American cultures, with performance on English data consistently surpassing that of other languages. This reveals a persistent cultural gap in LLMs, which complicates their ability to accurately process culturally rich and diverse figurative language such as proverbs. To address this, we introduce Jawaher, a benchmark designed to assess LLMs' capacity to comprehend and interpret Arabic proverbs. Jawaher includes proverbs from various Arabic dialects, along with idiomatic translations and explanations. Through extensive evaluations of both open- and closed-source models, we find that while LLMs can generate idiomatically accurate translations, they struggle with producing culturally nuanced and contextually relevant explanations. These findings highlight the need for ongoing model refinement and dataset expansion to bridge the cultural gap in figurative language processing.

Intelligent agents powered by large language models (LLMs) have recently demonstrated impressive capabilities and gained increasing popularity on social media platforms. While LLM agents are reshaping the ecology of social media, there exists a current gap in conducting a comprehensive evaluation of their ability to comprehend media content, understand user behaviors, and make intricate decisions. To address this challenge, we introduce SoMe, a pioneering benchmark designed to evaluate social media agents equipped with various agent tools for accessing and analyzing social media data. SoMe comprises a diverse collection of 8 social media agent tasks, 9,164,284 posts, 6,591 user profiles, and 25,686 reports from various social media platforms and external websites, with 17,869 meticulously annotated task queries. Compared with the existing datasets and benchmarks for social media tasks, SoMe is the first to provide a versatile and realistic platform for LLM-based social media agents to handle diverse social media tasks. By extensive quantitative and qualitative analysis, we provide the first overview insight into the performance of mainstream agentic LLMs in realistic social media environments and identify several limitations. Our evaluation reveals that both the current closed-source and open-source LLMs cannot handle social media agent tasks satisfactorily. SoMe provides a challenging yet meaningful testbed for future social media agents. Our code and data are available at https://github.com/LivXue/SoMe

Model distillation is a technique for transferring knowledge from large language models (LLMs) to smaller ones, aiming to create resource-efficient yet high-performing models. However, excessive distillation can lead to homogenization, reducing diversity among models and impairing their ability to robustly handle complex or novel tasks. These limitations underscore the need to systematically quantify the distillation process and its impact. In this work, we propose a framework to evaluate and quantify model distillation. Our method addresses two key aspects: (1) Identifying identity cognition contradictions to assess discrepancies in how models perceive and represent identity-related information, and (2) Analyzing multi-granularity response similarities across models to measure the extent of homogenization. Experimental results demonstrate two key insights: (1) Well-known closed-source and open-source LLMs usually exhibit high distillation degrees, except for Claude, Doubao, and Gemini. (2) Base LLMs show higher distillation degrees compared to aligned LLMs. By offering a systematic approach to improve the transparency of LLM data distillation, we call for LLMs with more independent development and more transparent technical reports to improve LLMs' robustness and safety. The code and data are available under https://github.com/Aegis1863/LLMs-Distillation-Quantification.

Hallucination is a persistent issue affecting all large language Models (LLMs), particularly within low-resource languages such as Persian. PerHalluEval (Persian Hallucination Evaluation) is the first dynamic hallucination evaluation benchmark tailored for the Persian language. Our benchmark leverages a three-stage LLM-driven pipeline, augmented with human validation, to generate plausible answers and summaries regarding QA and summarization tasks, focusing on detecting extrinsic and intrinsic hallucinations. Moreover, we used the log probabilities of generated tokens to select the most believable hallucinated instances. In addition, we engaged human annotators to highlight Persian-specific contexts in the QA dataset in order to evaluate LLMs' performance on content specifically related to Persian culture. Our evaluation of 12 LLMs, including open- and closed-source models using PerHalluEval, revealed that the models generally struggle in detecting hallucinated Persian text. We showed that providing external knowledge, i.e., the original document for the summarization task, could mitigate hallucination partially. Furthermore, there was no significant difference in terms of hallucination when comparing LLMs specifically trained for Persian with others.

There is a lack of benchmarks for evaluating large language models (LLMs) in long-form medical question answering (QA). Most existing medical QA evaluation benchmarks focus on automatic metrics and multiple-choice questions. While valuable, these benchmarks fail to fully capture or assess the complexities of real-world clinical applications where LLMs are being deployed. Furthermore, existing studies on evaluating long-form answer generation in medical QA are primarily closed-source, lacking access to human medical expert annotations, which makes it difficult to reproduce results and enhance existing baselines. In this work, we introduce a new publicly available benchmark featuring real-world consumer medical questions with long-form answer evaluations annotated by medical doctors. We performed pairwise comparisons of responses from various open and closed-source medical and general-purpose LLMs based on criteria such as correctness, helpfulness, harmfulness, and bias. Additionally, we performed a comprehensive LLM-as-a-judge analysis to study the alignment between human judgments and LLMs. Our preliminary results highlight the strong potential of open LLMs in medical QA compared to leading closed models. Code & Data: https://github.com/lavita-ai/medical-eval-sphere

Large language models (LLMs) have proven invaluable for code generation, particularly in interactive settings. However, existing code generation benchmarks fail to capture the diverse feedback encountered in multi-turn interactions, limiting our ability to evaluate LLMs in these contexts. To address this gap, we present a set of novel benchmarks that explicitly model the quality of feedback provided to code generation LLMs. Our contributions are threefold: First, we introduce CONVCODEWORLD, a novel and reproducible environment for benchmarking interactive code generation. CONVCODEWORLD simulates 9 distinct interactive code generation scenarios while systematically combining three types of feedback: (a) compilation feedback; (b) execution feedback with varying test coverage; (c) verbal feedback generated by GPT-4o with different levels of expertise. Second, we introduce CONVCODEBENCH, a fast, static version of benchmark that uses pre-generated feedback logs, eliminating the need for costly dynamic verbal feedback generation while maintaining strong Spearman's rank correlations (0.82 to 0.99) with CONVCODEWORLD. Third, extensive evaluations of both closed-source and open-source LLMs including R1-Distill on CONVCODEWORLD reveal key insights: (a) LLM performance varies significantly based on the feedback provided; (b) Weaker LLMs, with sufficient feedback, can outperform single-turn results of state-of-the-art LLMs without feedback; (c) Training on a specific feedback combination can limit an LLM's ability to utilize unseen combinations; (d) LLMs solve problems in fewer turns (high MRR) may not solve as many problems overall (high Recall), and vice versa. All implementations and benchmarks will be made publicly available at https://huggingface.co/spaces/ConvCodeWorld/ConvCodeWorld

Recent advances in agentic Large Language Models (LLMs) have positioned them as generalist planners capable of reasoning and acting across diverse tasks. However, existing agent benchmarks largely focus on symbolic or weakly grounded environments, leaving their performance in physics-constrained real-world domains underexplored. We introduce AstroReason-Bench, a comprehensive benchmark for evaluating agentic planning in Space Planning Problems (SPP), a family of high-stakes problems with heterogeneous objectives, strict physical constraints, and long-horizon decision-making. AstroReason-Bench integrates multiple scheduling regimes, including ground station communication and agile Earth observation, and provides a unified agent-oriented interaction protocol. Evaluating on a range of state-of-the-art open- and closed-source agentic LLM systems, we find that current agents substantially underperform specialized solvers, highlighting key limitations of generalist planning under realistic constraints. AstroReason-Bench offers a challenging and diagnostic testbed for future agentic research.

Large language models (LLMs) have shown increasing capability in problem-solving and decision-making, largely based on the step-by-step chain-of-thought reasoning processes. However, it has been increasingly challenging to evaluate the reasoning capability of LLMs. Concretely, existing outcome-based benchmarks begin to saturate and become less sufficient to monitor the progress. To this end, we present a process-based benchmark MR-BEN that demands a meta reasoning skill, where LMs are asked to locate and analyse potential errors in automatically generated reasoning steps. MR-BEN is a comprehensive benchmark comprising 5,975 questions collected from human experts, covering various subjects such as physics, chemistry, logic, coding, and more. Through our designed metrics for assessing meta-reasoning on this benchmark, we identify interesting limitations and weaknesses of current LLMs (open-source and closed-source models). For example, open-source models are seemingly comparable to GPT-4 on outcome-based benchmarks, but they lag far behind on our benchmark, revealing the underlying reasoning capability gap between them. Our dataset and codes are available on https://randolph-zeng.github.io/Mr-Ben.github.io/.

Low-resource languages face significant challenges due to the lack of sufficient linguistic data, resources, and tools for tasks such as supervised learning, annotation, and classification. This shortage hinders the development of accurate models and datasets, making it difficult to perform critical NLP tasks like sentiment analysis or hate speech detection. To bridge this gap, Large Language Models (LLMs) present an opportunity for potential annotators, capable of generating datasets and resources for these underrepresented languages. In this paper, we focus on Marathi, a low-resource language, and evaluate the performance of both closed-source and open-source LLMs as annotators. We assess models such as GPT-4o and Gemini 1.0 Pro, Gemma 2 (2B and 9B), and Llama 3.1 (8B) on classification tasks including sentiment analysis, news classification, and hate speech detection. Our findings reveal that while LLMs excel in annotation tasks for high-resource languages like English, they still fall short when applied to Marathi. Even advanced closed models like Gemini and GPT underperform in comparison to BERT-based baselines, highlighting the limitations of LLMs as annotators for low-resource languages.

Text-to-SQL, the task of translating natural language questions into SQL queries, plays a crucial role in enabling non-experts to interact with databases. While recent advancements in large language models (LLMs) have significantly enhanced text-to-SQL performance, existing approaches face notable limitations in real-world text-to-SQL applications. Prompting-based methods often depend on closed-source LLMs, which are expensive, raise privacy concerns, and lack customization. Fine-tuning-based methods, on the other hand, suffer from poor generalizability due to the limited coverage of publicly available training data. To overcome these challenges, we propose a novel and scalable text-to-SQL data synthesis framework for automatically synthesizing large-scale, high-quality, and diverse datasets without extensive human intervention. Using this framework, we introduce SynSQL-2.5M, the first million-scale text-to-SQL dataset, containing 2.5 million samples spanning over 16,000 synthetic databases. Each sample includes a database, SQL query, natural language question, and chain-of-thought (CoT) solution. Leveraging SynSQL-2.5M, we develop OmniSQL, a powerful open-source text-to-SQL model available in three sizes: 7B, 14B, and 32B. Extensive evaluations across nine datasets demonstrate that OmniSQL achieves state-of-the-art performance, matching or surpassing leading closed-source and open-source LLMs, including GPT-4o and DeepSeek-V3, despite its smaller size. We release all code, datasets, and models to support further research.

The capability gap between open-source and closed-source large language models (LLMs) remains a challenge in text-to-SQL tasks. In this paper, we introduce a synthetic data approach that combines data produced by larger, more powerful models (strong models) with error information data generated by smaller, not well-aligned models (weak models). The method not only enhances the domain generalization of text-to-SQL models but also explores the potential of error data supervision through preference learning. Furthermore, we employ the synthetic data approach for instruction tuning on open-source LLMs, resulting SENSE, a specialized text-to-SQL model. The effectiveness of SENSE is demonstrated through state-of-the-art results on the SPIDER and BIRD benchmarks, bridging the performance gap between open-source models and methods prompted by closed-source models.

This paper presents a novel approach for unified retrieval-augmented generation (RAG) systems using the recent emerging large language model (LLM) agent concept. Specifically, Agent LLM, which utilizes LLM as fundamental controllers, has become a promising approach to enable the interpretability of RAG tasks, especially for complex reasoning question-answering systems (e.g., multi-hop queries). Nonetheless, previous works mainly focus on solving RAG systems with either single-hop or multi-hop approaches separately, which limits the application of those approaches to real-world applications. In this study, we propose a trainable agent framework called Agent-UniRAG for unified retrieval-augmented LLM systems, which enhances the effectiveness and interpretability of RAG systems. The main idea is to design an LLM agent framework to solve RAG tasks step-by-step based on the complexity of the inputs, simultaneously including single-hop and multi-hop queries in an end-to-end manner. Furthermore, we introduce SynAgent-RAG, a synthetic dataset to enable the proposed agent framework for small open-source LLMs (e.g., Llama-3-8B). The results show comparable performances with closed-source and larger open-source LLMs across various RAG benchmarks. Our source code and dataset are publicly available for further exploitation.

Recently, Multimodal Large Language Models (MLLMs) have shown great promise in language-guided perceptual tasks such as recognition, segmentation, and object detection. However, their effectiveness in addressing visual cognition problems that require high-level reasoning is not well-established. One such challenge is abstract visual reasoning (AVR) -- the cognitive ability to discern relationships among patterns in a set of images and extrapolate to predict subsequent patterns. This skill is crucial during the early neurodevelopmental stages of children. Inspired by the AVR tasks in Raven's Progressive Matrices (RPM) and Wechsler Intelligence Scale for Children (WISC), we propose a new dataset MaRs-VQA and a new benchmark VCog-Bench containing three datasets to evaluate the zero-shot AVR capability of MLLMs and compare their performance with existing human intelligent investigation. Our comparative experiments with different open-source and closed-source MLLMs on the VCog-Bench revealed a gap between MLLMs and human intelligence, highlighting the visual cognitive limitations of current MLLMs. We believe that the public release of VCog-Bench, consisting of MaRs-VQA, and the inference pipeline will drive progress toward the next generation of MLLMs with human-like visual cognition abilities.

Recently, both closed-source LLMs and open-source communities have made significant strides, outperforming humans in various general domains. However, their performance in specific professional fields such as medicine, especially within the open-source community, remains suboptimal due to the complexity of medical knowledge. We propose Aquila-Med, a bilingual medical LLM based on Aquila, addressing these challenges through continue pre-training, supervised fine-tuning (SFT), and reinforcement learning from human feedback (RLHF). We construct a large-scale Chinese and English medical dataset for continue pre-training and a high-quality SFT dataset, covering extensive medical specialties. Additionally, we develop a high-quality Direct Preference Optimization (DPO) dataset for further alignment. Aquila-Med achieves notable results across single-turn, multi-turn dialogues, and medical multiple-choice questions, demonstrating the effectiveness of our approach. We open-source the datasets and the entire training process, contributing valuable resources to the research community. Our models and datasets will released at https://huggingface.co/BAAI/AquilaMed-RL.

Recently, both closed-source LLMs and open-source communities have made significant strides, outperforming humans in various general domains. However, their performance in specific professional domains such as medicine, especially within the open-source community, remains suboptimal due to the complexity of medical knowledge. In this paper, we propose CareBot, a bilingual medical LLM, which leverages a comprehensive approach integrating continuous pre-training (CPT), supervised fine-tuning (SFT), and reinforcement learning with human feedback (RLHF). Our novel two-stage CPT method, comprising Stable CPT and Boost CPT, effectively bridges the gap between general and domain-specific data, facilitating a smooth transition from pre-training to fine-tuning and enhancing domain knowledge progressively. We also introduce DataRater, a model designed to assess data quality during CPT, ensuring that the training data is both accurate and relevant. For SFT, we develope a large and diverse bilingual dataset, along with ConFilter, a metric to enhance multi-turn dialogue quality, which is crucial to improving the model's ability to handle more complex dialogues. The combination of high-quality data sources and innovative techniques significantly improves CareBot's performance across a range of medical applications. Our rigorous evaluations on Chinese and English benchmarks confirm CareBot's effectiveness in medical consultation and education. These advancements not only address current limitations in medical LLMs but also set a new standard for developing effective and reliable open-source models in the medical domain. We will open-source the datasets and models later, contributing valuable resources to the research community.

Recent advancements in large language models (LLMs) have significantly advanced text-to-SQL systems. However, most LLM-based methods often narrowly focus on SQL generation, neglecting the complexities of real-world conversational queries. This oversight can lead to unreliable responses, particularly for ambiguous questions that cannot be directly addressed with SQL. To bridge this gap, we propose MMSQL, a comprehensive test suite designed to evaluate the question classification and SQL generation capabilities of LLMs by simulating real-world scenarios with diverse question types and multi-turn Q&A interactions. Using MMSQL, we assessed the performance of popular LLMs, including both open-source and closed-source models, and identified key factors impacting their performance in such scenarios. Moreover, we introduce an LLM-based multi-agent framework that employs specialized agents to identify question types and determine appropriate answering strategies. Our experiments demonstrate that this approach significantly enhances the model's ability to navigate the complexities of conversational dynamics, effectively handling the diverse and complex nature of user queries. Our dataset and code are publicly available at https://mcxiaoxiao.github.io/MMSQL.

The advent of large language models (LLMs) has unlocked great opportunities in complex data management tasks, particularly in question answering (QA) over complicated multi-table relational data. Despite significant progress, systematically evaluating LLMs on multi-table QA remains a critical challenge due to the inherent complexity of analyzing heterogeneous table structures and potential large scale of serialized relational data. Existing benchmarks primarily focus on single-table QA, failing to capture the intricacies of reasoning across multiple relational tables, as required in real-world domains such as finance, healthcare, and e-commerce. To address this gap, we present TQA-Bench, a new multi-table QA benchmark designed to evaluate the capabilities of LLMs in tackling complex QA tasks over relational data. Our benchmark incorporates diverse relational database instances sourced from real-world public datasets and introduces a flexible sampling mechanism to create tasks with varying multi-table context lengths, ranging from 8K to 64K tokens. To ensure robustness and reliability, we integrate symbolic extensions into the evaluation framework, enabling the assessment of LLM reasoning capabilities beyond simple data retrieval or probabilistic pattern matching. We systematically evaluate a range of LLMs, both open-source and closed-source, spanning model scales from 7 billion to 70 billion parameters. Our extensive experiments reveal critical insights into the performance of LLMs in multi-table QA, highlighting both challenges and opportunities for advancing their application in complex, data-driven environments. Our benchmark implementation and results are available at https://github.com/Relaxed-System-Lab/TQA-Bench.

Teaching language models to use search tools is not only a question of whether they search, but also of whether they issue good queries. This is especially important in open-domain question answering, where broad or copied queries often waste retrieval budget and derail later reasoning. We propose \Ours, a framework that makes query planning explicit through reusable search skills. At each step, the model first selects a skill, then generates a search or answer action conditioned on the selected skill card. The skill inventory itself is not fixed: SearchSkill maintains an evolving SkillBank, expands or refines it from recurrent failure patterns, and reconstructs affected trajectories before supervised training. The resulting two-stage SFT recipe aligns training with the inference-time protocol of skill selection followed by skill-grounded execution. Across open-source and closed-source models, SearchSkill improves exact match on knowledge-intensive QA benchmarks and yields better retrieval behavior, including fewer copied first queries, more atomic hop-focused queries, and more correct answers within a small search budget. These results suggest that explicit skill-conditioned query planning is a lightweight alternative to treating search as an undifferentiated action.

Retrieval-augmented generation (RAG) has emerged as a promising solution to mitigate the limitations of large language models (LLMs), such as hallucinations and outdated information. However, it remains unclear how LLMs handle knowledge conflicts arising from different augmented retrieved passages, especially when these passages originate from the same source and have equal trustworthiness. In this work, we conduct a comprehensive evaluation of LLM-generated answers to questions that have varying answers based on contradictory passages from Wikipedia, a dataset widely regarded as a high-quality pre-training resource for most LLMs. Specifically, we introduce WikiContradict, a benchmark consisting of 253 high-quality, human-annotated instances designed to assess LLM performance when augmented with retrieved passages containing real-world knowledge conflicts. We benchmark a diverse range of both closed and open-source LLMs under different QA scenarios, including RAG with a single passage, and RAG with 2 contradictory passages. Through rigorous human evaluations on a subset of WikiContradict instances involving 5 LLMs and over 3,500 judgements, we shed light on the behaviour and limitations of these models. For instance, when provided with two passages containing contradictory facts, all models struggle to generate answers that accurately reflect the conflicting nature of the context, especially for implicit conflicts requiring reasoning. Since human evaluation is costly, we also introduce an automated model that estimates LLM performance using a strong open-source language model, achieving an F-score of 0.8. Using this automated metric, we evaluate more than 1,500 answers from seven LLMs across all WikiContradict instances. To facilitate future work, we release WikiContradict on: https://ibm.biz/wikicontradict.

This paper introduces UnSeenTimeQA, a novel time-sensitive question-answering (TSQA) benchmark that diverges from traditional TSQA benchmarks by avoiding factual and web-searchable queries. We present a series of time-sensitive event scenarios decoupled from real-world factual information. It requires large language models (LLMs) to engage in genuine temporal reasoning, disassociating from the knowledge acquired during the pre-training phase. Our evaluation of six open-source LLMs (ranging from 2B to 70B in size) and three closed-source LLMs reveal that the questions from the UnSeenTimeQA present substantial challenges. This indicates the models' difficulties in handling complex temporal reasoning scenarios. Additionally, we present several analyses shedding light on the models' performance in answering time-sensitive questions.

The advent of Multimodal Large Language Models (MLLMs) has expanded AI capabilities to visual modalities, yet existing evaluation benchmarks remain limited to single-video understanding, overlooking the critical need for multi-video understanding in real-world scenarios (e.g., sports analytics and autonomous driving). To address this significant gap, we introduce MVU-Eval, the first comprehensive benchmark for evaluating Multi-Video Understanding for MLLMs. Specifically, our MVU-Eval mainly assesses eight core competencies through 1,824 meticulously curated question-answer pairs spanning 4,959 videos from diverse domains, addressing both fundamental perception tasks and high-order reasoning tasks. These capabilities are rigorously aligned with real-world applications such as multi-sensor synthesis in autonomous systems and cross-angle sports analytics. Through extensive evaluation of state-of-the-art open-source and closed-source models, we reveal significant performance discrepancies and limitations in current MLLMs' ability to perform understanding across multiple videos. The benchmark will be made publicly available to foster future research.

In this work, we introduce a novel evaluation paradigm for Large Language Models, one that challenges them to engage in meta-reasoning. This approach addresses critical shortcomings in existing math problem-solving benchmarks, traditionally used to evaluate the cognitive capabilities of agents. Our paradigm shifts the focus from result-oriented assessments, which often overlook the reasoning process, to a more holistic evaluation that effectively differentiates the cognitive capabilities among models. For example, in our benchmark, GPT-4 demonstrates a performance ten times more accurate than GPT3-5. The significance of this new paradigm lies in its ability to reveal potential cognitive deficiencies in LLMs that current benchmarks, such as GSM8K, fail to uncover due to their saturation and lack of effective differentiation among varying reasoning abilities. Our comprehensive analysis includes several state-of-the-art math models from both open-source and closed-source communities, uncovering fundamental deficiencies in their training and evaluation approaches. This paper not only advocates for a paradigm shift in the assessment of LLMs but also contributes to the ongoing discourse on the trajectory towards Artificial General Intelligence (AGI). By promoting the adoption of meta-reasoning evaluation methods similar to ours, we aim to facilitate a more accurate assessment of the true cognitive abilities of LLMs.

In this study, we focus on two main tasks, the first for detecting legal violations within unstructured textual data, and the second for associating these violations with potentially affected individuals. We constructed two datasets using Large Language Models (LLMs) which were subsequently validated by domain expert annotators. Both tasks were designed specifically for the context of class-action cases. The experimental design incorporated fine-tuning models from the BERT family and open-source LLMs, and conducting few-shot experiments using closed-source LLMs. Our results, with an F1-score of 62.69\% (violation identification) and 81.02\% (associating victims), show that our datasets and setups can be used for both tasks. Finally, we publicly release the datasets and the code used for the experiments in order to advance further research in the area of legal natural language processing (NLP).

Effective deep search agents must not only access open-domain and domain-specific knowledge but also apply complex rules-such as legal clauses, medical manuals and tariff rules. These rules often feature vague boundaries and implicit logic relationships, making precise application challenging for agents. However, this critical capability is largely overlooked by current agent benchmarks. To fill this gap, we introduce HSCodeComp, the first realistic, expert-level e-commerce benchmark designed to evaluate deep search agents in hierarchical rule application. In this task, the deep reasoning process of agents is guided by these rules to predict 10-digit Harmonized System Code (HSCode) of products with noisy but realistic descriptions. These codes, established by the World Customs Organization, are vital for global supply chain efficiency. Built from real-world data collected from large-scale e-commerce platforms, our proposed HSCodeComp comprises 632 product entries spanning diverse product categories, with these HSCodes annotated by several human experts. Extensive experimental results on several state-of-the-art LLMs, open-source, and closed-source agents reveal a huge performance gap: best agent achieves only 46.8% 10-digit accuracy, far below human experts at 95.0%. Besides, detailed analysis demonstrates the challenges of hierarchical rule application, and test-time scaling fails to improve performance further.