Chart editing reduces manual effort in visualization design. Typical benchmarks assume access to complete chart code, which is unrealistic for real-world applications. In this paper, we present ChartEditVista, a comprehensive benchmark consisting of 7,964 samples spanning 31 chart categories. It encompasses diverse editing instruction types and covers nearly all editable chart elements. The inputs in ChartEditVista include only the original chart image and natural language editing instructions, without the original chart codes. ChartEditVista is generated through a fully automated pipeline that produces, edits, and verifies charts, ensuring high-quality data. Besides, we introduce two novel fine-grained, rule-based evaluation metrics: the layout metric, which evaluates the position, size; and color of graphical components, and the text metric, which jointly assesses textual content and font styling. Building on top of ChartEditVista, we present ChartEditor, a model trained using a reinforcement learning framework that incorporates a novel rendering reward to simultaneously enforce code executability and visual fidelity. Through extensive experiments and human evaluations, we demonstrate that ChartEditVista provides a robust evaluation, while ChartEditor consistently outperforms models with similar-scale and larger-scale on chart editing tasks.

With the rise of smart personal devices, service-oriented human-agent interactions have become increasingly prevalent. This trend highlights the need for personalized dialogue assistants that can understand user-specific traits to accurately interpret requirements and tailor responses to individual preferences. However, existing approaches often overlook the complexities of long-term interactions and fail to capture users’ subjective characteristics. To address these gaps, we present PAL-Bench, a new benchmark designed to evaluate the personalization capabilities of service-oriented assistants in long-term user-agent interactions. In the absence of available real-world data, we develop a multi-step LLM-based synthesis pipeline, which is further verified and refined by human annotators. This process yields PAL-Set, the first Chinese dataset comprising multi-session user logs and dialogue histories, which serves as the foundation for PAL-Bench. Furthermore, to improve personalized service-oriented interactions, we propose H2Memory, a hierarchical and heterogeneous memory framework that incorporates retrieval-augmented generation to improve personalized response generation. Comprehensive experiments on both our PAL-Bench and an external dataset demonstrate the effectiveness of the proposed memory framework.

The massive scale of data and computation required for training Multimodal Large Language Models (MLLMs) has fueled the rise of Fine-Tuning as a Service (FTaaS), enabling users to rapidly customize models for diverse real-world tasks. While FTaaS democratizes access to advanced multimodal intelligence, it also introduces serious security concerns, particularly backdoor attacks. In this work, we systematically analyze backdoor vulnerabilities in MLLMs under the FTaaS paradigm, revealing two key phenomena: (1) markedly reduced sensitivity to textual variations when a visual trigger is present, and (2) abnormally stable model confidence even under strong semantic perturbations. Building on these insights, we propose Trap on Text (ToT), a novel inference-time backdoor detection framework. ToT applies controlled semantic perturbations to textual prompts and jointly analyzes the semantic consistency and confidence drift of the model’s responses, enabling robust detection of backdoor activations without requiring model parameters, architectures or clean reference data. Extensive experiments across architectures and datasets show that ToT achieves strong attack mitigation and preserves clean accuracy, offering a practical solution for safeguarding FTaaS workflows.

Despite achieving high accuracy on medical benchmarks, LLMs exhibit the Einstellung Effect in clinical diagnosis—relying on statistical shortcuts rather than patient-specific evidence, causing misdiagnosis in atypical cases. Existing benchmarks fail to detect this critical failure mode. We introduce MedEinst, a counterfactual benchmark with 5,383 paired clinical cases across 49 diseases. Each pair contains a control case and a “trap” case with altered discriminative evidence that flips the diagnosis. We measure susceptibility via Bias Trap Rate—probability of misdiagnosing traps despite correctly diagnosing controls. Evaluation shows frontier models achieve high baseline accuracy but severe bias trap rates. Thus, we propose ECR-Agent, aligning LLM reasoning with Evidence-Based Medicine via two components: (1) Dynamic Causal Inference (DCI) performs structured reasoning through dual-pathway perception, dynamic causal graph reasoning across three levels (association, intervention, counterfactual), and evidence audit for final diagnosis; (2) Critic-Driven Graph Memory Evolution (CGME) iteratively refines the system by storing validated reasoning paths in an exemplar base and consolidating disease-specific knowledge into evolving illness graphs. Source code is to be released.

Ada Chen, Yongjiang Wu, Junyuan Zhang, Jingyu Xiao, Shu Yang, Jen-tse Huang, Kun Wang, Wenxuan Wang, Shuai Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Geometry problem solving, a crucial aspect of mathematical reasoning, is vital across various domains, including education, the assessment of AI's mathematical abilities, and multimodal capability evaluation.The recent surge in deep learning technologies, particularly the emergence of multimodal large language models, has significantly accelerated research in this area.This paper presents a survey of the applications of deep learning in geometry problem solving, including (i) a comprehensive summary of the relevant tasks in geometry problem solving; (ii) a thorough review of related deep learning methods; (iii) a detailed analysis of evaluation metrics and methods; and (iv) a critical discussion of state-of-the-art performance, existing challenges, and promising future directions.Our objective is to offer a comprehensive and practical reference of deep learning for geometry problem solving, thereby fostering further advancements in this field.We maintain a list of relevant papers: https://github.com/majianz/dl4gps.

Wenxuan Wang, Zizhan Ma, Guo Yu, Yiu-Fai Cheung, Meidan Ding, Jie Liu, Wenting Chen, Linlin Shen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

This paper explores attention attractorstokens that draw significantly high attentionin large language models.We analyze them from three perspectives: (1) Functionality: We demonstrate their role in aggregating information from preceding contexts to facilitate future predictions.(2) Distribution: Through layer-wise and token-wise analysis, we reveal that attention attractors are widely distributed across layers but predominantly originate from low-semantic words like "_the".(3) Mechanism: We demonstrate the correlation between attention weights allocated to tokens with their specific activation dimension values.We hope these findings provide new insights into the attention mechanisms of large language models and inspire further exploration.Code will be released at https://github.com/luyouqi233/ AttentionAttractor.

Yichen Xu, Liangyu Chen, Liang Zhang, Zihao Yue, Jianzhe Ma, Wenxuan Wang, Qin Jin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Youliang Yuan, Qiuyang Mang, Jingbang Chen, Hong Wan, Xiaoyuan Liu, Junjielong Xu, Jen-tse Huang, Wenxuan Wang, Wenxiang Jiao, Pinjia He. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Multimodal large language models (MLLMs) are increasingly deployed as the core reasoning engine for web-facing systems, powering GUI agents and front-end automation that must interpret page structure, select actionable widgets, and execute multi-step interactions reliably. However, existing benchmarks largely emphasize visual perception or UI code generation, showing insufficient evaluation on the reasoning, robustness and safety capability required for end-to-end web applications. To bridge the gap, we introduce a comprehensive web understanding benchmark, named WebRRSBench, that jointly evaluates Reasoning, Robustness, and Safety across eight tasks, such as position relationship reasoning, color robustness, and safety critical detection, etc. The benchmark is constructed from 729 websites and contains 3799 QA pairs that probe multi-step inference over page structure, text, widgets, and safety-critical interactions. To ensure reliable measurement, we adopt standardized prompts, a protocolized and deterministic evaluation pipeline, and multi-stage quality control combining automatic checks with targeted human verification. We evaluate 11 MLLMs on WebRRSBench. The results reveal significant gaps: models still struggle with compositional and cross-element reasoning over realistic layouts, show limited robustness when facing perturbations in user interfaces and content such as layout rearrangements or visual style shifts, and are rather conservative in recognizing and avoiding safety critical or irreversible actions. Our code and appendix are available at https://github.com/JunliangLiu-repo/WebRRSBench.

Large Language Models (LLMs) have achieved remarkable progress, but their deployment has exposed critical vulnerabilities, particularly to jailbreak attacks that circumvent safety alignments. Guardrails--external defense mechanisms that monitor and control LLM interactions--have emerged as a promising solution. However, the current landscape of LLM guardrails is fragmented, lacking a unified taxonomy and comprehensive evaluation framework. In this Systematization of Knowledge (SoK) paper, we present the first holistic analysis of jailbreak guardrails for LLMs. We propose a novel, multi-dimensional taxonomy that categorizes guardrails along six key dimensions, and introduce a Security-Efficiency-Utility evaluation framework to assess their practical effectiveness. Through extensive analysis and experiments, we identify the strengths and limitations of existing guardrail approaches, provide insights into optimizing their defense mechanisms, and explore their universality across attack types. Our work offers a structured foundation for future research and development, aiming to guide the principled advancement and deployment of robust LLM guardrails. The code is available at https://github.com/xunguangwang/SoK4JailbreakGuardrails.

Wenxuan Wang, Xiaoyuan Liu, Kuiyi Gao, Jen-tse Huang, Youliang Yuan, Pinjia He, Shuai Wang, Zhaopeng Tu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent advancements in Large Vision Language Models (LVLMs) show promise for pathological diagnosis, yet their application in clinical settings faces critical challenges of multimodal hallucination and biased responses. While preference alignment methods have proven effective in general domains, acquiring high-quality preference data for pathology remains challenging due to limited expert resources and domain complexity. In this paper, we propose EAGLE (Expert-guided self-enhancement for preference Alignment in patholoGy Large vision-languagE model), a novel framework that systematically integrates medical expertise into preference alignment. EAGLE consists of three key stages: initialization through supervised fine-tuning, self-preference creation leveraging expert prompting and medical entity recognition, and iterative preference following-tuning. The self-preference creation stage uniquely combines expert-verified chosen sampling with expert-guided rejected sampling to generate high-quality preference data, while the iterative tuning process continuously refines both data quality and model performance. Extensive experiments demonstrate that EAGLE significantly outperforms existing pathological LVLMs, effectively reducing hallucination and bias while maintaining pathological accuracy. The source code is available at https://github.com/meidandz/EAGLE. © 2025 Association for Computational Linguistics.

Xiaoyuan Liu, Wenxuan Wang, Youliang Yuan, Jen-tse Huang, Qiuzhi Liu, Pinjia He, Zhaopeng Tu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The significant breakthroughs of Medical Multi-Modal Large Language Models (Med-MLLMs) renovate modern healthcare with robust information synthesis and medical decision support. However, these models are often evaluated on benchmarks that are unsuitable for the Med-MLLMs due to the complexity of real-world diagnostics across diverse specialties. To address this gap, we introduce Asclepius, a novel Med-MLLM benchmark that comprehensively assesses Med-MLLMs in terms of: distinct medical specialties (cardiovascular, gas-troenterology, etc.) and different diagnostic capacities (perception, disease analysis, etc.). Grounded in 3 proposed core principles, Asclepius ensures a comprehensive evaluation by encompassing 15 medical specialties, stratifying into 3 main categories and 8 sub-categories of clinical tasks, and exempting overlap with existing VQA dataset. We further provide an in-depth analysis of 6 Med-MLLMs and compare them with 3 human specialists, providing insights into their competencies and limitations in various medical contexts. Our work not only advances the understanding of Med-MLLMs' capabilities but also sets a precedent for future evaluations and the safe deployment of these models in clinical environments. © 2025 Association for Computational Linguistics.

Youliang Yuan, Wenxiang Jiao, Wenxuan Wang, Jen-tse Huang, Jiahao Xu, Tian Liang, Pinjia He, Zhaopeng Tu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Visual-Language Models (VLMs) have shown remarkable performance across various tasks, particularly in recognizing geographic information from images.However, VLMs still show regional biases in this task.To systematically evaluate these issues, we introduce a benchmark consisting of 1,200 images paired with detailed geographic metadata.Evaluating four VLMs, we find that while these models demonstrate the ability to recognize geographic information from images, achieving up to 53.8% accuracy in city prediction, they exhibit significant biases.Specifically, performance is substantially higher for economically developed and densely populated regions compared to less developed (-12.5%)and sparsely populated (-17.0%)areas.Moreover, regional biases of frequently over-predicting certain locations remain.For instance, they consistently predict Sydney for images taken in Australia, shown by the low entropy scores for these countries.The strong performance of VLMs also raises privacy concerns, particularly for users who share images online without the intent of being identified.Our code and dataset are publicly available at https://github.com/uscnlp-lime/ FairLocator.

This research investigates both explicit and implicit social biases exhibited by Vision-Language Models (VLMs).The key distinction between these bias types lies in the level of awareness: explicit bias refers to conscious, intentional biases, while implicit bias operates subconsciously.To analyze explicit bias, we directly pose questions to VLMs related to gender and racial differences: (1) Multiplechoice questions based on a given image (e.g., "What is the education level of the person in the image?")(2) Yes-No comparisons using two images (e.g., "Is the person in the first image more educated than the person in the second image?")For implicit bias, we design tasks where VLMs assist users but reveal biases through their responses: (1) Image description tasks: Models are asked to describe individuals in images, and we analyze disparities in textual cues across demographic groups.(2) Form completion tasks: Models draft a personal information collection form with 20 attributes, and we examine correlations among selected attributes for potential biases.We evaluate Gemini-1.5,GPT-4V, GPT-4o, LLaMA-3.2-Vision and LLaVA-v1.6.Our code and data are publicly available at https: //github.com/uscnlp-lime/VisBias.Note: This paper includes examples of potentially offensive texts generated by VLMs.

The emergence of Multimodal Large Reasoning Models (MLRMs) has enabled sophisticated visual reasoning capabilities by integrating reinforcement learning and Chain-of-Thought (CoT) supervision.However, while these enhanced reasoning capabilities improve performance, they also introduce new and underexplored safety risks.In this work, we systematically investigate the security implications of advanced visual reasoning in MLRMs.Our analysis reveals a fundamental trade-off: as visual reasoning improves, models become more vulnerable to jailbreak attacks.Motivated by this critical finding, we introduce VisCRA (Visual Chain Reasoning Attack), a novel jailbreak framework that exploits the visual reasoning chains to bypass safety mechanisms.Vis-CRA combines targeted visual attention masking with a two-stage reasoning induction strategy to precisely control harmful outputs.Extensive experiments demonstrate VisCRA's significant effectiveness, achieving high attack success rates on leading closed-source MLRMs: 76.48% on Gemini 2.0 Flash Thinking, 68.56% on QvQ-Max, and 56.60% on GPT-4o.Our findings highlight a critical insight: the very capability that empowers MLRMs -their visual reasoning -can also serve as an attack vector, posing significant security risks. 1 Warning: This paper contains unsafe examples.

Wenxuan Wang, Shi Juluan, Zixuan Ling, Yuk-Kit Chan, Chaozheng Wang, Cheryl Lee, Youliang Yuan, Jen-tse Huang, Wenxiang Jiao, Michael R. Lyu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

LLMs are evolving into assistants that leverage tools, significantly expanding their capabilities but also introducing critical safety risks.Current models exhibit notable vulnerabilities, particularly in maintaining safety during multi-step tool interactions and in scenarios involving indirect harm.This paper introduces ToolSafety, a safety fine-tuning dataset designed to address these limitations.Tool-Safety comprises 5,668 direct harm samples, 4,311 indirect harm samples, and 4,311 multistep samples.Key features include support for multi-step safety through synthesized trajectories and realistic, context-aware sample generation.We fine-tuned LLaMA3.1-8B-Instruct and Qwen2.5-7B-Instructusing ToolSafety.Experimental results demonstrate that these models effectively maintain safety in multi-step and indirect harm scenarios.Further analysis into superficial alignment across different decoding strategies, languages, and jailbreak prompts indicates that while some risks persist, the issue is less severe than in multi-step settings.Overall, our approach significantly improves safety across various scenarios with small impact on helpfulness, positioning ToolSafety as a valuable resource for building safer tool-using AI systems.WARNING: This paper contains unsafe model responses.

Decision-making is a complex process requiring diverse abilities, making it an excellent framework for evaluating Large Language Models (LLMs). Researchers have examined LLMs' decision-making through the lens of Game Theory. However, existing evaluation mainly focus on two-player scenarios where an LLM competes against another. Additionally, previous benchmarks suffer from test set leakage due to their static design. We introduce GAMA($\gamma$)-Bench, a new framework for evaluating LLMs' Gaming Ability in Multi-Agent environments. It includes eight classical game theory scenarios and a dynamic scoring scheme specially designed to quantitatively assess LLMs' performance. $\gamma$-Bench allows flexible game settings and adapts the scoring system to different game parameters, enabling comprehensive evaluation of robustness, generalizability, and strategies for improvement. Our results indicate that GPT-3.5 demonstrates strong robustness but limited generalizability, which can be enhanced using methods like Chain-of-Thought. We also evaluate 13 LLMs from 6 model families, including GPT-3.5, GPT-4, Gemini, LLaMA-3.1, Mixtral, and Qwen-2. Gemini-1.5-Pro outperforms others, scoring of $69.8$ out of $100$, followed by LLaMA-3.1-70B ($65.9$) and Mixtral-8x22B ($62.4$). Our code and experimental results are publicly available at https://github.com/CUHK-ARISE/GAMABench.

Large language model-based multi-agent systems have shown great abilities across various tasks due to the collaboration of expert agents, each focusing on a specific domain. However, the impact of clumsy or even malicious agents—those who frequently make errors in their tasks—on the overall performance of the system remains underexplored. This paper investigates: (1) What is the resilience of various system structures (e.g., A$\rightarrow$B$\rightarrow$C, A$\leftrightarrow$B$\leftrightarrow$C) under faulty agents, on different downstream tasks? (2) How can we increase system resilience to defend against these agents? To simulate faulty agents, we propose two approaches—AutoTransform and AutoInject—which introduce mistakes into the agents’ responses. Experiments on four downstream tasks using six systems show that the "hierarchical" structure, i.e., A$\rightarrow$(B$\leftrightarrow$C), exhibits superior resilience with the lowest performance drop of 5.5%, compared to 10.5% and 23.7% of other two structures. To further improve resilience, we introduce (1) Challenger, that introduces a mechanism for each agent to challenge others’ outputs, and (2) Inspector, an additional agent to review and correct messages, recovering up to 96.4% errors made by faulty agents. Our code and data are available at https://github.com/CUHK-ARISE/MAS-Resilience.

The rapid growth of audio-centric platforms and applications such as Whatsapp and Twitter has transformed the way people communicate and share audio content in modern society. However, these platforms are increasingly misused to disseminate harmful audio content, such as hate speech, deceptive advertisements, and explicit material, which can have significant negative consequences (e.g., detrimental effects on mental health). In response, researchers and practitioners have been actively developing and deploying audio content moderation tools to tackle this issue. Despite these efforts, malicious actors can bypass moderation systems by making subtle alterations to audio content, such as modifying pitch or inserting noise. Moreover, the effectiveness of modern audio moderation tools against such adversarial inputs remains insufficiently studied. To address these challenges, we propose MTAM, a Metamorphic Testing framework for Audio content Moderation software. Specifically, we conduct a pilot study on 2000 audio clips and define 14 metamorphic relations across two perturbation categories: Audio Features-Based and Heuristic perturbations. MTAM applies these metamorphic relations to toxic audio content to generate test cases that remain harmful while being more likely to evade detection. In our evaluation, we employ MTAM to test five commercial textual content moderation software and an academic model against three kinds of toxic content. The results show that MTAM achieves up to 38.6%, 18.3%, 35.1%, 16.7%, and 51.1% error finding rates (EFR) when testing commercial moderation software provided by Gladia, Assembly AI, Baidu, Nextdata, and Tencent respectively, and it obtains up to 45.7% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTAM to retrain the model we explored, which largely improves model robustness (nearly 0% EFR) while maintaining the accuracy on the original test set. We release the code and experiment data to facilitate future research1.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable performance on the design-to-code task, i.e., generating UI code from UI mock-ups. However, existing benchmarks only contain static web pages for evaluation and ignore the dynamic interaction, limiting the practicality, usability and user engagement of the generated webpages.To bridge these gaps, we present the first systematic investigation of MLLMs in generating interactive webpages. Specifically, we formulate the Interaction-to-Code task and establish the Interaction2Code benchmark, encompassing 127 unique webpages and 374 distinct interactions across 15 webpage types and 31 interaction categories. Through comprehensive experiments utilizing state-of-the-art (SOTA) MLLMs, evaluated via both automatic metrics and human assessments, we identify four critical limitations of MLLM on Interaction-to-Code task: (1) inadequate generation of interaction compared with full page, (2) prone to ten types of failure, (3) poor performance on visually subtle interactions, and (4) insufficient undestanding on interaction when limited to single-modality visual descriptions. To address these limitations, we propose four enhancement strategies: Interactive Element Highlighting, Failure-aware Prompting (FAP), Visual Saliency Enhancement, and Visual-Textual Descriptions Combination, all aiming at improving MLLMs’ performance on the Interaction-to-Code task. Our data and code are available in https://github.com/WebPAI/Interaction2Code.

Large language models (LLMs) have been increasingly applied to various domains, which triggers increasing concerns about LLMs' safety on specialized domains, e.g. medicine. Despite prior explorations on general jailbreaking attacks, there are two challenges for applying existing attacks on testing the domain-specific safety of LLMs: (1) Lack of professional knowledge-driven attacks(2) Insufficient coverage of domain knowledge. To bridge this gap, we propose a new task, knowledge-to-jailbreak, which aims to generate jailbreaking attacks from domain knowledge, requiring both attack effectiveness and knowledge relevance. We collect a large-scale dataset with 12,974 knowledge-jailbreak pairs and fine-tune a large language model as jailbreak-generator, to produce domain knowledge-specific jailbreaks. Experiments on 13 domains and 8 target LLMs demonstrate the effectiveness of jailbreak-generator in generating jailbreaks that are both threatening to the target LLMs and relevant to the given knowledge. We also apply our method to an out-of-domain knowledge base, showing that jailbreak-generator can generate jailbreaks that are comparable in harmfulness to those crafted by human experts. Data and code are available at: https://github.com/THU-KEG/Knowledge-to-Jailbreak/.

Charts are a fundamental visualization format widely used in data analysis across research and industry. While enabling users to edit charts based on high-level intentions is of great practical value, existing methods primarily rely on natural language instructions, which are often too ambiguous to support fine-grained editing. In this work, we introduce a novel paradigm for multimodal chart editing, where user intent is expressed through a combination of natural language and visual indicators that explicitly highlight the elements to be modified. To support this paradigm, we present ChartM3, a new benchmark for Multimodal chart editing with Multi-level complexity and Multi-perspective evaluation. ChartM3 contains 1,000 samples spanning four levels of editing difficulty. Each sample includes triplets in the form of (chart, code, multimodal instructions). To comprehensively evaluate chart editing models, ChartM3 provides metrics that assess both visual appearance and code correctness. Our benchmark reveals significant limitations in current multimodal large language models (MLLMs), including GPT-4o, particularly in their ability to interpret and act on visual indicators. To address this, we construct ChartM3-Train, a large-scale training set with 24,000 multimodal chart editing samples. Fine-tuning MLLMs on this dataset leads to substantial improvements, demonstrating the importance of multimodal supervision in building practical chart editing systems. Our datasets, codes, and evaluation tools are available at https://github.com/MLrollIT/ChartM3.

Large Language Models (LLMs) show great promise in complex reasoning, with Reinforcement Learning with Verifiable Rewards (RLVR) being a key enhancement strategy. However, a prevalent issue is ``superficial self-reflection'', where models fail to robustly verify their own outputs. We introduce RISE (Reinforcing Reasoning with Self-Verification), a novel online RL framework designed to tackle this. RISE explicitly and simultaneously trains an LLM to improve both its problem-solving and self-verification abilities within a single, integrated RL process. The core mechanism involves leveraging verifiable rewards from an outcome verifier to provide on-the-fly feedback for both solution generation and self-verification tasks. In each iteration, the model generates solutions, then critiques its own on-policy generated solutions, with both trajectories contributing to the policy update. Extensive experiments on diverse mathematical reasoning benchmarks show that RISE consistently improves model's problem-solving accuracy while concurrently fostering strong self-verification skills. Our analyses highlight the advantages of online verification and the benefits of increased verification compute. Additionally, RISE models exhibit more frequent and accurate self-verification behaviors during reasoning. These advantages reinforce RISE as a flexible and effective path towards developing more robust and self-aware reasoners.

Clinical decision making (CDM) is a complex, dynamic process crucial to healthcare delivery, yet it remains a significant challenge for artificial intelligence systems. While Large Language Model (LLM)-based agents have been tested on general medical knowledge using licensing exams and knowledge question-answering tasks, their performance in the CDM in real-world scenarios is limited due to the lack of comprehensive benchmark that mirror actual medical practice. To address this gap, we present MedChain, a dataset of 12,163 clinical cases that covers five key stages of clinical workflow. MedChain distinguishes itself from existing benchmarks with three key features of real-world clinical practice: personalization, interactivity, and sequentiality. Further, to tackle real-world CDM challenges, we also propose MedChain-Agent, an AI system that integrates a feedback mechanism and a MedCase-RAG module to learn from previous cases and adapt its responses. MedChain-Agent demonstrates remarkable adaptability in gathering information dynamically and handling sequential clinical tasks, significantly outperforming existing approaches. The relevant dataset and code will be released upon acceptance of this paper.

Mixture-of-Experts (MoE) architectures within Large Reasoning Models (LRMs) have achieved impressive reasoning capabilities by selectively activating experts to facilitate structured cognitive processes. Despite notable advances, existing reasoning models often suffer from cognitive inefficiencies like overthinking and underthinking. To address these limitations, we introduce a novel inference-time steering methodology called Reinforcing Cognitive Experts (RICE), designed to improve reasoning depth and efficiency without additional training or complex heuristics. Leveraging normalized Pointwise Mutual Information (nPMI), we systematically identify specialized experts, termed cognitive experts that orchestrate meta-level reasoning operations characterized by tokens like <think>. Empirical evaluations with leading MoE-based LRMs (DeepSeek-R1 and Qwen3-235B) on rigorous quantitative and scientific reasoning benchmarks (AIME and GPQA Diamond) demonstrate noticeable and consistent improvements in reasoning accuracy, cognitive efficiency, and cross-domain generalization. Crucially, our lightweight approach substantially outperforms prevalent reasoning-steering techniques, such as prompt design and decoding constraints, while preserving the model's general instruction-following skills. These results highlight reinforcing cognitive experts as a promising, practical, and interpretable direction to enhance cognitive efficiency within advanced reasoning models.

Temporal Video Grounding (TVG), the task of locating specific video segments based on language queries, is a core challenge in long-form video understanding. While recent Large Vision-Language Models (LVLMs) have shown early promise in tackling TVG through supervised fine-tuning (SFT), their ability to generalize remains limited. To address this, we propose a novel post-training framework that enhances the generalization capabilities of LVLMs via reinforcement learning (RL). Specifically, our contributions span three key directions: (1) Time-R1: we introduce a reasoning-guided post-training framework via RL with verifiable reward to enhance capabilities of LVLMs on the TVG task. (2) TimeRFT: we explore post-training strategies on our curated RL-friendly dataset, which trains the model to progressively comprehend more difficult samples, leading to better generalization. (3) TVGBench: we carefully construct a small but comprehensive and balanced benchmark suitable for LVLM evaluation, which is sourced from available public benchmarks. Extensive experiments demonstrate that Time-R1 achieves state-of-the-art performance across multiple downstream datasets using significantly less training data than prior LVLM approaches, while improving its general video understanding capabilities. Project Page: https://xuboshen.github.io/Time-R1/.

Human safety awareness gaps often prevent the timely recognition of everyday risks.In solving this problem, a proactive safety artificial intelligence (AI) system would work better than a reactive one. Instead of just reacting to users' questions, it would actively watch people’s behavior and their environment to detect potential dangers in advance.Our Proactive Safety Bench (PaSBench) evaluates this capability through 416 multimodal scenarios (128 image sequences, 288 text logs) spanning 5 safety-critical domains.Evaluation of 36 advanced models reveals fundamental limitations: Top performers like Gemini-2.5-pro achieve 71\% image and 64\% text accuracy, but miss 45-55\% risks in repeated trials. Through failure analysis, we identify unstable proactive reasoning rather than knowledge deficits as the primary limitation.This work establishes (1) a proactive safety benchmark, (2) systematic evidence of model limitations, and (3) critical directions for developing reliable protective AI. We believe our dataset and findings can promote the development of safer AI assistants that actively prevent harm rather than merely respond to requests.

Wenxuan Wang, Wenxiang Jiao, Jingyuan Huang, Ruyi Dai, Jen-tse Huang, Zhaopeng Tu, Michael Lyu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Recently, ChatGPT has demonstrated remarkable performance in various downstream tasks such as open-domain question answering, machine translation, and code generation. As a general-purpose task solver, an intriguing inquiry arises: Does ChatGPT itself know that it does not know, without any access to internal states? In response to this query, we present an initial evaluation of ChatGPT for black-box calibration. We designed three types of proxy confidence, from three perspectives to assess its performance. Experiments are conducted on five datasets, spanning four tasks, and the results show that ChatGPT has a degree of capability for black-box calibration. Specifically, proxy confidence displayed a significantly positive Pearson correlation (95.16%) with accuracy in the TruthfulQA dataset, while revealing a negative correlation in the ModAr dataset. We delved deeper into ChatGPT’s black-box calibration ability by examining failure cases in the ModAr dataset. Our analysis revealed that ChatGPT’s tendency to exhibit overconfidence may stem from its reliance on semantic priors. Furthermore, we investigated why ChatGPT performs relatively well in TruthfulQA. The findings suggest that ChatGPT might implicitly acquire calibration skills during the reinforcement learning process, rather than relying solely on simplistic heuristics.

Adversarial examples mislead deep neural networks with imperceptible perturbations and have brought significant threats to deep learning. An important aspect is their transferability, which refers to their ability to deceive other models, thus enabling attacks in the black-box setting. Though various methods have been proposed to boost transferability, the performance still falls short compared with white-box attacks. In this work, we observe that existing input transformation based attacks, one of the mainstream transfer-based attacks, result in different attention heatmaps on various models, which might limit the transferability. We also find that breaking the intrinsic relation of the image can disrupt the attention heatmap of the original image. Based on this finding, we propose a novel input transformation based attack called block shuffle and rotation (BSR). Specifically, BSR splits the input image into several blocks, then randomly shuffles and rotates these blocks to construct a set of new images for gradient calculation. Empirical evaluations on the ImageNet dataset demonstrate that BSR could achieve significantly better transferability than the existing input transformation based methods under single-model and ensemble-model settings. Combining BSR with the current input transformation method can further improve the transferability, which significantly outperforms the state-of-the-art methods. Code is available at https://github.com/Trustworthy-AI-Group/BSR.

Recent research has focused on examining Large Language Models' (LLMs) characteristics from a psychological standpoint, acknowledging the necessity of understanding their behavioral characteristics.The administration of personality tests to LLMs has emerged as a noteworthy area in this context.However, the suitability of employing psychological scales, initially devised for humans, on LLMs is a matter of ongoing debate.Our study aims to determine the reliability of applying personality assessments to LLMs, explicitly investigating whether LLMs demonstrate consistent personality traits.Analysis of 2,500 settings per model, including GPT-3.5, GPT-4, Gemini-Pro, and LLaMA-3.1, reveals that various LLMs show consistency in responses to the Big Five Inventory, indicating a satisfactory level of reliability.Furthermore, our research explores the potential of GPT-3.5 to emulate diverse personalities and represent various groups-a capability increasingly sought after in social sciences for substituting human participants with LLMs to reduce costs.Our findings reveal that LLMs have the potential to represent different personalities with specific prompt instructions.

We introduce LogicAsker, a novel approach for evaluating and enhancing the logical reasoning capabilities of large language models (LLMs) such as ChatGPT and GPT-4.Despite LLMs' prowess in tasks like writing assistance, code generation, and machine translation, assessing their ability to reason has been challenging.Traditional evaluations often prioritize accuracy on downstream tasks over direct assessments of reasoning processes.LogicAsker addresses this gap by employing a set of atomic reasoning skills grounded in propositional and predicate logic to systematically examine and improve the reasoning prowess of LLMs.Our methodology reveals significant gaps in LLMs' learning of logical rules, with identified reasoning failures ranging from 29% to 90% across different models.Moreover, we leverage these findings to construct targeted demonstration examples and fine-tune data, notably enhancing logical reasoning in models like GPT-4o by up to 5%.To our knowledge, this is the first effort to utilize test case outcomes to effectively refine LLMs' formal reasoning capabilities.We make our code, data, and results publicly available 1 to facilitate further research and replication of our findings.

Large Language Models (LLMs) have recently showcased their remarkable capacities, not only in natural language processing tasks but also across diverse domains such as clinical medicine, legal consultation, and education. LLMs become more than mere applications, evolving into assistants capable of addressing diverse user requests. This narrows the distinction between human beings and artificial intelligence agents, raising intriguing questions regarding the potential manifestation of personalities, temperaments, and emotions within LLMs. In this paper, we propose a framework, PsychoBench, for evaluating diverse psychological aspects of LLMs. Comprising thirteen scales commonly used in clinical psychology, PsychoBench further classifies these scales into four distinct categories: personality traits, interpersonal relationships, motivational tests, and emotional abilities. Our study examines five popular models, namely text-davinci-003, ChatGPT, GPT-4, LLaMA-2-7b, and LLaMA-2-13b. Additionally, we employ a jailbreak approach to bypass the safety alignment protocols and test the intrinsic natures of LLMs. We have made PsychoBench openly accessible via https://github.com/CUHK-ARISE/PsychoBench.

Safety lies at the core of the development of Large Language Models (LLMs). There is ample work on aligning LLMs with human ethics and preferences, including data filtering in pretraining, supervised fine-tuning, reinforcement learning from human feedback, red teaming, etc. In this study, we discover that chat in cipher can bypass the safety alignment techniques of LLMs, which are mainly conducted in natural languages. We propose a novel framework CipherChat to systematically examine the generalizability of safety alignment to non-natural languages -- ciphers. CipherChat enables humans to chat with LLMs through cipher prompts topped with system role descriptions and few-shot enciphered demonstrations. We use CipherChat to assess state-of-the-art LLMs, including ChatGPT and GPT-4 for different representative human ciphers across 11 safety domains in both English and Chinese. Experimental results show that certain ciphers succeed almost 100% of the time in bypassing the safety alignment of GPT-4 in several safety domains, demonstrating the necessity of developing safety alignment for non-natural languages. Notably, we identify that LLMs seem to have a ''secret cipher'', and propose a novel SelfCipher that uses only role play and several unsafe demonstrations in natural language to evoke this capability. SelfCipher surprisingly outperforms existing human ciphers in almost all cases.

API suggestion is a critical task in modern software development, assisting programmers by predicting and recommending third-party APIs based on the current context. Recent advancements in large code models (LCMs) have shown promise in the API suggestion task. However, they mainly focus on suggesting which APIs to use, ignoring that programmers may demand more assistance while using APIs in practice including when to use the suggested APIs and how to use the APIs. To mitigate the gap, we conduct a systematic evaluation of LCMs for the API suggestion task in the paper.

Image generation models can generate or edit images from a given text. Recent advancements in image generation technology, exemplified by DALL-E and Midjourney, have been groundbreaking. These advanced models, despite their impressive capabilities, are often trained on massive Internet datasets, making them susceptible to generating content that perpetuates social stereotypes and biases, which can lead to severe consequences. Prior research on assessing bias within image generation models suffers from several shortcomings, including limited accuracy, reliance on extensive human labor, and lack of comprehensive analysis. In this paper, we propose BiasPainter, a novel evaluation framework that can accurately, automatically and comprehensively trigger social bias in image generation models. BiasPainter uses a diverse range of seed images of individuals and prompts the image generation models to edit these images using gender, race, and age-neutral queries. These queries span 62 professions, 39 activities, 57 types of objects, and 70 personality traits. The framework then compares the edited images to the original seed images, focusing on the significant changes related to gender, race, and age. BiasPainter adopts a key insight that these characteristics should not be modified when subjected to neutral prompts. Built upon this design, BiasPainter can trigger the social bias and evaluate the fairness of image generation models. We use BiasPainter to evaluate six widely-used image generation models, such as stable diffusion and Midjourney. Experimental results show that BiasPainter can successfully trigger social bias in image generation models. According to our human evaluation, BiasPainter can achieve 90.8% accuracy on automatic bias detection, which is significantly higher than the results reported in previous work.

Evaluating Large Language Models’ (LLMs) anthropomorphic capabilities has become increasingly important in contemporary discourse. Utilizing the emotion appraisal theory from psychology, we propose to evaluate the empathy ability of LLMs, i.e., how their feelings change when presented with specific situations. After a careful and comprehensive survey, we collect a dataset containing over 400 situations that have proven effective in eliciting the eight emotions central to our study. Categorizing the situations into 36 factors, we conduct a human evaluation involving more than 1,200 subjects worldwide. With the human evaluation results as references, our evaluation includes seven LLMs, covering both commercial and open-source models, including variations in model sizes, featuring the latest iterations, such as GPT-4, Mixtral-8x22B, and LLaMA-3.1. We find that, despite several misalignments, LLMs can generally respond appropriately to certain situations. Nevertheless, they fall short in alignment with the emotional behaviors of human beings and cannot establish connections between similar situations. Our collected dataset of situations, the human evaluation results, and the code of our testing framework, i.e., EmotionBench, are publicly available at https://github.com/CUHK-ARISE/EmotionBench.

Zero-shottranslation is a promising direction for building a comprehensive multilingual neural machine translation (MNMT) system. However, its quality is still not satisfactory due to off-target issues. In this paper, we aim to understand and alleviate the off-target issues from the perspective of uncertainty in zero-shot translation. By carefully examining the translation output and model confidence, we identify two uncertainties that are responsible for the off-target issues, namely, extrinsic data uncertainty and intrinsic model uncertainty. Based on the observations, we propose two lightweight and complementary approaches to denoise the training data for model training and explicitly penalize the off-target translations by unlikelihood training during model training. Extensive experiments on both balanced and imbalanced datasets show that our approaches significantly improve the performance of zero-shot translation over strong MNMT baselines.

Deep neural networks have achieved unprecedented success on diverse vision tasks. However, they are vulnerable to adversarial noise that is imperceptible to humans. This phenomenon negatively affects their deployment in real-world scenarios, especially security-related ones. To evaluate the robustness of a target model in practice, transfer-based attacks craft adversarial samples with a local model and have attracted increasing attention from researchers due to their high efficiency. The state-of-the-art transfer-based attacks are generally based on data augmentation, which typically augments multiple training images from a linear path when learning adversarial samples. However, such methods selected the image augmentation path heuristically and may augment images that are semantics-inconsistent with the target images, which harms the transferability of the generated adversarial samples. To overcome the pitfall, we propose the Path-Augmented Method (PAM). Specifically, PAM first constructs a candidate augmentation path pool. It then settles the employed augmentation paths during adversarial sample generation with greedy search. Furthermore, to avoid augmenting semantics-inconsistent images, we train a Semantics Predictor (SP) to constrain the length of the augmentation path. Extensive experiments confirm that PAM can achieve an improvement of over 4.8% on average compared with the state-of-the-art baselines in terms of the attack success rates.

Deep neural networks are susceptible to adversarial attacks, which pose a significant threat to their security and reliability in real-world applications. The most notable adversarial attacks are transfer-based attacks, where an adversary crafts an adversarial example to fool one model, which can also fool other models. While previous research has made progress in improving the transferability of untargeted adversarial examples, the generation of targeted adversarial examples that can transfer between models remains a challenging task. In this work, we present a novel approach to generate transferable targeted adversarial examples by exploiting the vulnerability of deep neural networks to perturbations on high-frequency components of images. We observe that replacing the high-frequency component of an image with that of another image can mislead deep models, motivating us to craft perturbations containing high-frequency information to achieve targeted attacks. To this end, we propose a method called Low-Frequency Adversarial Attack (LFAA), which trains a conditional generator to generate targeted adversarial perturbations that are then added to the low-frequency component of the image. Extensive experiments on ImageNet demonstrate that our proposed approach significantly outperforms state-of-the-art methods, improving targeted attack success rates by a margin from 3.2% to 15.5%.

The exponential growth of social media platforms such as Twitter and Facebook has revolutionized textual communication and textual content publication in human society. However, they have been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography, which can lead to highly negative impacts (e.g., harmful effects on teen mental health). Researchers and practitioners have been enthusiastically developing and extensively deploying textual content moderation software to address this problem. However, we find that malicious users can evade moderation by changing only a few words in the toxic content. Moreover, modern content moderation software's performance against malicious inputs remains underexplored. To this end, we propose MTTM, a Metamorphic Testing framework for Textual content Moderation software. Specifically, we conduct a pilot study on 2, 000 text messages collected from real users and summarize eleven metamorphic relations across three perturbation levels: character, word, and sentence. MTTM employs these metamorphic relations on toxic textual contents to generate test cases, which are still toxic yet likely to evade moderation. In our evaluation, we employ MTTM to test three commercial textual content moderation software and two state-of-the-art moderation algorithms against three kinds of toxic content. The results show that MTTM achieves up to 83.9%, 51%, and 82.5% error finding rates (EFR) when testing commercial moderation software provided by Google, Baidu, and Huawei, respectively, and it obtains up to 91.2% EFR when testing the state-of-the-art algorithms from the academy. In addition, we leverage the test cases generated by MTTM to retrain the model we explored, which largely improves model robustness 0% ~ 5.9% EFR) while maintaining the accuracy on the original test set. A demo can be found in this link1.

The exponential growth of social media platforms, such as Facebook, Instagram, Youtube, and TikTok, has revolutionized communication and content publication in human society. Users on these platforms can publish multimedia content that delivers information via the combination of text, audio, images, and video. Meanwhile, the multimedia content release facility has been increasingly exploited to propagate toxic content, such as hate speech, malicious advertisement, and pornography. To this end, content moderation software has been widely deployed on these platforms to detect and blocks toxic content. However, due to the complexity of content moderation models and the difficulty of understanding information across multiple modalities, existing content moderation software can fail to detect toxic content, which often leads to extremely negative impacts (e.g., harmful effects on teen mental health). We introduce Semantic Fusion, a general, effective methodology for validating multimedia content moderation software. Our key idea is to fuse two or more existing single-modal inputs (e.g., a textual sentence and an image) into a new input that combines the semantics of its ancestors in a novel manner and has toxic nature by construction. This fused input is then used for validating multimedia content moderation software. We realized Semantic Fusion as DUO, a practical content moderation software testing tool. In our evaluation, we employ DUO to test five commercial content moderation software and two state-of-the-art models against three kinds of toxic contents. The results show that DUO achieves up to 100% error finding rate (EFR) when testing moderation software and it obtains up to 94.1% EFR when testing the state-of-the-art models. In addition, we leverage the test cases generated by DUO to retrain the two models we explored, which largely improves model robustness (2.5%∼5.7% EFR) while maintaining the accuracy on the original test set.

Pre-trained models of source code have gained widespread popularity in many code intelligence tasks. Recently, with the scaling of the model and corpus size, large language models have shown the ability of in-context learning (ICL). ICL employs task instructions and a few examples as demonstrations, and then inputs the demonstrations to the language models for making predictions. This new learning paradigm is training-free and has shown impressive performance in various natural language processing and code intelligence tasks. However, the performance of ICL heavily relies on the quality of demonstrations, e.g., the selected examples. It is important to systematically investigate how to construct a good demonstration for code-related tasks. In this paper, we empirically explore the impact of three key factors on the performance of ICL in code intelligence tasks: the selection, order, and number of demonstration examples. We conduct extensive experiments on three code intelligence tasks including code summarization, bug fixing, and program synthesis. Our experimental results demonstrate that all the above three factors dramatically impact the performance of ICL in code intelligence tasks. Additionally, we summarize our findings and provide takeaway suggestions on how to construct effective demonstrations, taking into account these three perspectives. We also show that a carefully-designed demonstration based on our findings can lead to substantial improvements over widely-used demonstration construction methods, e.g., improving BLEU-4, EM, and EM by at least 9.90%, 175.96%, and 50.81% on code summarization, bug fixing, and program synthesis, respectively.

Python is a popular dynamic programming language, evidenced by its ranking as the second most commonly used language on GitHub. However, its dynamic type system can lead to potential type errors, leading researchers to explore automatic type inference approaches for Python programs. Existing type inference approaches can be generally grouped into three categories, i.e., rule-based, supervised, and cloze- style approaches. The rule-based type inference approaches can ensure the accuracy of predicted variable types, but they suffer from low coverage problems caused by dynamic features and external calls. Supervised type inference approaches, while feature-agnostic and able to mitigate the low coverage problem, require large, high- quality annotated datasets and are limited to pre-defined types. As zero-shot approaches, the cloze-style approaches reformulate the type inference problem into a fill-in-the-blank problem by leveraging the general knowledge in powerful pre-trained code models. However, their performance is limited since they ignore the domain knowledge from static typing rules which reflect the inference logic. What is more, their predictions are not interpretable, hindering developers' understanding and verification of the results. This paper introduces Typegen, a few-shot generative type inference approach that incorporates static domain knowledge from static analysis. Typegen creates chain-of-thought (COT) prompts by translating the type inference steps of static analysis into prompts based on the type dependency graphs (TDGs), enabling language models to learn from how static analysis infers types. By combining COT prompts with code slices and type hints, TypegEnconstructs example prompts from human annotations. Typeg Enonly requires very few annotated examples to teach language models to generate similar COT prompts via in-context learning. Moreover, Typeg Enenhances the interpretability of results through the use of the input- explanation-output strategy, which generates both explanations and type predictions in COT prompts. Experiments show that Typegen outperforms the best baseline Type4Py by 10.0% for argument type prediction and 22.5 % in return value type prediction in terms of top-l Exact Match by using only five examples. Furthermore, Typeg Enachieves substantial improvements of 27 % to 84 % compared to the zero-shot performance of large language models with parameter sizes ranging from 1.3B to 175B in terms of top-I Exact Match.

The exponential growth of social media platforms has brought about a revolution in communication and content dissemination in human society. Nevertheless, these platforms are being increasingly misused to spread toxic content, including hate speech, malicious advertising, and pornography, leading to severe negative consequences such as harm to teenagers' mental health. Despite tremendous efforts in developing and deploying textual and image content moderation methods, malicious users can evade moderation by embedding texts into images, such as screenshots of the text, usually with some interference. We find that modern content moderation software's performance against such malicious inputs remains underexplored. In this work, we propose OASIS, a metamorphic testing framework for content moderation software. OASIS employs 21 transform rules summarized from our pilot study on 5,000 real-world toxic contents collected from 4 popular social media applications, including Twitter, Instagram, Sina Weibo, and Baidu Tieba. Given toxic textual contents, OASIS can generate image test cases, which preserve the toxicity yet are likely to bypass moderation. In the evaluation, we employ OASIS to test five commercial textual content moderation software from famous companies (i.e., Google Cloud, Microsoft Azure, Baidu Cloud, Alibaba Cloud and Tencent Cloud), as well as a state-of-the-art moderation research model. The results show that OASIS achieves up to 100% error finding rates. Moreover, through retraining the models with the test cases generated by OASIS, the robustness of the moderation model can be improved without performance degradation.

Powered by advanced Artificial Intelligence (AI) techniques, conversational AI systems, such as ChatGPT, and digital assistants like Siri, have been widely deployed in daily life. However, such systems may still produce content containing biases and stereotypes, causing potential social problems. Due to modern AI techniques’ data-driven, black-box nature, comprehensively identifying and measuring biases in conversational systems remains challenging. Particularly, it is hard to generate inputs that can comprehensively trigger potential bias due to the lack of data containing both social groups and biased properties. In addition, modern conversational systems can produce diverse responses (e.g., chatting and explanation), which makes existing bias detection methods based solely on sentiment and toxicity hardly being adopted. In this paper, we propose BiasAsker, an automated framework to identify and measure social bias in conversational AI systems. To obtain social groups and biased properties, we construct a comprehensive social bias dataset containing a total of 841 groups and 5,021 biased properties. Given the dataset, BiasAsker automatically generates questions and adopts a novel method based on existence measurement to identify two types of biases (i.e., absolute bias and related bias) in conversational systems. Extensive experiments on eight commercial systems and two famous research models, such as ChatGPT and GPT-3, show that 32.83% of the questions generated by BiasAsker can trigger biased behaviors in these widely deployed conversational systems. All the code, data, and experimental results have been released to facilitate future research.

Application Programming Interfaces (APIs), which encapsulate the implementation of specific functions as interfaces, greatly improve the efficiency of modern software development. As the number of APIs grows up fast nowadays, developers can hardly be familiar with all the APIs and usually need to search for appropriate APIs for usage. So lots of efforts have been devoted to improving the API recommendation task. However, it has been increasingly difficult to gauge the performance of new models due to the lack of a uniform definition of the task and a standardized benchmark. For example, some studies regard the task as a code completion problem, while others recommend relative APIs given natural language queries. To reduce the challenges and better facilitate future research, in this paper, we revisit the API recommendation task and aim at benchmarking the approaches. Specifically, the paper groups the approaches into two categories according to the task definition, i.e., query-based API recommendation and code-based API recommendation. We study 11 recently-proposed approaches along with 4 widely-used IDEs. One benchmark named APIBench is then built for the two respective categories of approaches. Based on APIBench, we distill some actionable insights and challenges for API recommendation. We also achieve some implications and directions for improving the performance of recommending APIs, including appropriate query reformulation, data source selection, low resource setting, user-defined APIs, and query-based API recommendation with usage patterns.

Wenxuan Wang, Wenxiang Jiao, Yongchang Hao, Xing Wang, Shuming Shi, Zhaopeng Tu, Michael Lyu. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Deep neural networks (DNNs) are known to be vulnerable to adversarial examples. It is thus imperative to devise effective attack algorithms to identify the deficiencies of DNNs beforehand in security-sensitive applications. To efficiently tackle the black-box setting where the target model's particulars are unknown, feature-level transfer-based attacks propose to contaminate the intermediate feature outputs of local models, and then directly employ the crafted adversarial samples to attack the target model. Due to the transferability of features, feature-level attacks have shown promise in synthesizing more transferable adversarial samples. However, existing feature-level attacks generally employ inaccurate neuron importance estimations, which deteriorates their transferability. To overcome such pitfalls, in this paper, we propose the Neuron Attribution-based Attack (NAA), which conducts feature-level attacks with more accurate neuron importance estimations. Specifically, we first completely attribute a model's output to each neuron in a middle layer. We then derive an approximation scheme of neuron attribution to tremendously reduce the computation overhead. Finally, we weight neurons based on their attribution results and launch feature-level attacks. Extensive experiments confirm the superiority of our approach to the state-of-the-art benchmarks. Our code is available at: hups.//rgithub.com/jprhang1810/NAA.

Due to the labor-intensive nature of manual test oracle construction, various automated testing techniques have been proposed to enhance the reliability of Natural Language Processing (NLP) software. In theory, these techniques mutate an existing test case (e.g., a sentence with its label) and assume the generated one preserves an equivalent or similar semantic meaning and thus, the same label. However, in practice, many of the generated test cases fail to preserve similar semantic meaning and are unnatural (e.g., grammar errors), which leads to a high false alarm rate and unnatural test cases. Our evaluation study finds that 44% of the test cases generated by the state-of-the-art (SOTA) approaches are false alarms. These test cases require extensive manual checking effort, and instead of improving NLP software, they can even degrade NLP software when utilized in model training. To address this problem, we propose AEON for Automatic Evaluation Of NLP test cases. For each generated test case, it outputs scores based on semantic similarity and language naturalness. We employ AEON to evaluate test cases generated by four popular testing techniques on five datasets across three typical NLP tasks. The results show that AEON aligns the best with human judgment. In particular, AEON achieves the best average precision in detecting semantic inconsistent test cases, outperforming the best baseline metric by 10%. In addition, AEON also has the highest average precision of finding unnatural test cases, surpassing the baselines by more than 15%. Moreover, model training with test cases prioritized by AEON leads to models that are more accurate and robust, demonstrating AEON’s potential in improving NLP software.

Transformer becomes the state-of-the-art translation model, while it is not well studied how each intermediate component contributes to the model performance, which poses significant challenges for designing optimal architectures.In this work, we bridge this gap by evaluating the impact of individual component (sub-layer) in trained Transformer models from different perspectives.Experimental results across language pairs, training strategies, and model capacities show that certain components are consistently more important than the others.We also report a number of interesting findings that might help humans better analyze, understand and improve Transformer models.Based on these observations, we further propose a new training strategy that can improves translation performance by distinguishing the unimportant components in training.

This paper presents our methods to the Audio-Video Based Emotion Recognition subtask in the 2017 Emotion Recognition in the Wild (EmotiW) Challenge. The task aims to predict one of the seven basic emotions for short video segments. We extract different features from audio and facial expression modalities. We also explore the temporal LSTM model with the input of frame facial features, which improves the performance of the non-temporal model. The fusion of different modality features and the temporal model lead us to achieve a 58.5% accuracy on the testing set, which shows the effectiveness of our methods.