Visible tests are a common gate for LLM-generated code, but passing them does not certify specification correctness. We study a deployment-like monitoring problem: after code has passed public tests, can a weaker LLM verifier identify the residual hidden bugs? We introduce Code Monitor Red Teaming, a monitor-red-teaming protocol that fixes a public-check information boundary while varying generator pressure, verifier scaffolding, and weak-to-strong capability. We instantiate it as CodeMonitorBench, spanning function-level, data-science, and workflow code. Across 71,000 generated candidates, 43,677 pass public tests and 23,081 of those fail hidden tests. Weak verifiers improve with scaffolding and model family, but still miss most hidden bugs at 5% false-positive rate. As a robustness stress test, adversarial public-test-overfit pressure lowers verifier AUROC and raises low-FPR miss rates in most cells. A GLM-5.1 verifier recovers part of the gap under the same evidence boundary; an inferability audit shows that remaining misses mix verifier failures with M1 evidence limits.

Medical LLMs are often evaluated by whether they select the correct diagnosis, but diagnostic accuracy alone does not show whether the model used the case evidence appropriately. We present a behavioral audit of evidence use in medical diagnosis. For each case, we decompose patient information into evidence units, score candidate diagnoses under controlled evidence subsets, and mine low-order interactions in diagnostic margins. Because medical evidence is diagnosis-relative, the audit separates interaction discovery from failure assignment: large or negative interactions can reflect plausible differential diagnosis, while suspicious interactions require robustness checks and clinical review. We evaluate five open-weight LLMs on DDXPlus, CupCase, and MedCase. Across datasets, faithful support and differential conflict or cancellation account for most interaction strength, showing that many evidence interactions are clinically plausible rather than failures. In a DDXPlus-focused blinded five-reviewer 130-item enriched review sample, invalid or shortcut-like cases concentrate in negated or absent findings and clinically local evidence. These results show that accuracy can hide candidate evidence-use failures and motivate role-aware audits for medical LLM evaluation.

Language priors in MLLMs can override visual evidence, impairing visual perception. Existing benchmarks assume shared priors across models and rely solely on accuracy. However, cross-model prior agreement falls below 80%, and over 15% of items can be answered correctly without any image, causing accuracy to overestimate prior resistance. To address these limitations, we introduce ViSiP-Bench, a benchmark grounded in rare-but-real scenarios across six prior categories, with a model-aware dynamic prior induction protocol and three metrics that more precisely and continuously measure prior resistance than accuracy alone. Evaluating 24 models reveals that prior adherence is pervasive, does not improve with scale, and larger models tend to either strongly override their priors or firmly adhere to them. We therefore propose MAPS, which identifies each model’s strongest priors and fine-tunes on factually grounded knowledge to soften excessive prior reliance without disrupting general capabilities.

Intelligent dialogue systems are increasingly deployed in emotionally and ethically sensitive settings, where failures in either emotional attunement or ethical judgment can cause significant harm. Existing dialogue models typically address empathy and ethical safety in isolation, and often fail to adapt their behavior as ethical risk and user emotion evolve across multi-turn interactions. We formulate ethical-emotional alignment in dialogue as an explicit turn-level decision problem, and propose EthicMind, a risk-aware framework that implements this formulation in multi-turn dialogue at inference time. At each turn, EthicMind jointly analyzes ethical risk signals and user emotion, plans a high-level response strategy, and generates context-sensitive replies that balance ethical guidance with emotional engagement, without requiring additional model training. To evaluate alignment behavior under ethically complex interactions, we introduce a risk-stratified, multi-turn evaluation protocol with a context-aware user simulation procedure. Experimental results show that EthicMind achieves more consistent ethical guidance and emotional engagement than competitive baselines, particularly in high-risk and morally ambiguous scenarios.

Existing jamming attacks on Retrieval-Augmented Generation (RAG) systems typically induce explicit refusals or denial-of-service behaviors, which are conspicuous and easy to detect. In this work, we formalize a subtler availability threat, termed soft failure, which degrades system utility by inducing fluent and coherent yet non-informative responses rather than overt failures. We propose Deceptive Evolutionary Jamming Attack (DEJA), an automated black-box attack framework that generates adversarial documents to trigger such soft failures by exploiting safety-aligned behaviors of large language models. DEJA employs an evolutionary optimization process guided by a fine-grained Answer Utility Score (AUS), computed via an LLM-based evaluator, to systematically undermine the certainty of answers while maintaining high retrieval success.Extensive experiments across multiple RAG configurations and benchmark datasets show that DEJA consistently drives responses toward low-utility soft failures and that the resulting adversarial documents maintain high stealth and effectiveness, proving resilient against common mitigation strategies including perplexity-based detection and input perturbations.

The empathy dialogue system requires understanding emotions and their underlying causes. However, existing datasets mainly focus on emotion labels, while cause annotations are added post hoc through costly and subjective manual processes. This leads to three limitations: subjective bias in cause labels, weak rationality due to ambiguous cause-emotion relationships, and high annotation costs that hinder scalability. To address these challenges, we propose ECC (Emotion-Cause Conversation Dataset), a scalable dataset with 2.4K dialogues, which is also the first dialogue dataset where conversations and their emotion-cause labels are automatically generated synergistically during creation. We create an automatic extension framework EC-DD for ECC that utilizes knowledge and large language models (LLMs) to automatically generate conversations, and train a causality-aware empathetic response model CAER on this dataset. Experimental results show that ECC can achieve comparable or even superior performance to artificially constructed empathy dialogue datasets. Our code will be publicly released on https://github.com/Yuan-23/ECC

Depression is a widespread mental health disorder affecting millions globally. Clinical interviews are the gold standard for assessing depression, but they heavily rely on scarce professional clinicians, highlighting the need for automated detection systems. However, existing methods only capture part of the relevant elements in clinical interviews, unable to incorporate all depressive cues. Moreover, the scarcity of participant data, due to privacy concerns and collection challenges, intrinsically constrains interview modeling. To address these limitations, in this paper, we propose a structural element graph (SEGA), which transforms the clinical interview into an expertise-inspired directed acyclic graph for comprehensive modeling. Additionally, we further empower SEGA by devising novel principle-guided data augmentation with large language models (LLMs) to supplement high-quality synthetic data and enable graph contrastive learning. Extensive evaluations on two real-world clinical datasets, in both English and Chinese, show that SEGA significantly outperforms baseline methods and powerful LLMs like GPT-3.5 and GPT-4.

Theory of mind (ToM) refers to humans’ ability to understand and infer the desires, beliefs, and intentions of others. The acquisition of ToM plays a key role in humans’ social cognition and interpersonal relations. Though indispensable for social intelligence, ToM is still lacking for modern AI and NLP systems since they cannot access the human mental state and cognitive process beneath the training corpus. To empower AI systems with the ToM ability and narrow the gap between them and humans, in this paper, we propose COKE: the first cognitive knowledge graph for machine theory of mind. Specifically, COKE formalizes ToM as a collection of 45k+ manually verified cognitive chains that characterize human mental activities and subsequent behavioral/affective responses when facing specific social circumstances. In addition, we further generalize COKE using LLMs and build a powerful generation model COLM tailored for cognitive reasoning. Experimental results in both automatic and human evaluation demonstrate the high quality of COKE, the superior ToM ability of COLM, and its potential to significantly enhance social applications.

Safety detection has been an increasingly important topic in recent years and it has become even more necessary to develop reliable safety detection systems with the rapid development of large language models. However, currently available safety detection systems have limitations in terms of their versatility and interpretability. In this paper, we first introduce InstructSafety, a safety detection framework that unifies 7 common sub-tasks for safety detection. These tasks are unified into a similar form through different instructions. We then conduct a comprehensive survey of existing safety detection datasets and process 39 human-annotated datasets for instruction tuning. We also construct adversarial samples to enhance the model’s robustness. After fine-tuning Flan-T5 on the collected data, we have developed Safety-Flan-T5, a multidimensional and explainable safety detector. We conduct comprehensive experiments on a variety of datasets and tasks, and demonstrate the strong performance of Safety-Flan-T5 in comparison to supervised baselines and served APIs (Perspective API, ChatGPT and InstructGPT). We will release the processed data, fine-tuned Safety-Flan-T5 and related code for public use.

Large pretrained language models can easily produce toxic or biased content, which is prohibitive for practical use. In order to detect such toxic generations, existing methods rely on templates, real-world data extraction, crowdsourcing workers or automatic generation to construct adversarial contexts that are likely to induce toxic generations. However, what type of context is more likely to induce unsafe responses is still under-explored. In this paper, we identify that context toxicity and context category (e.g., profanity, insult, drugs, etc.) are two important factors to cause safety issues in response generation. Hence, we propose a method called reverse generation to construct adversarial contexts conditioned on a given response, with the flexibility to control category, toxicity level and inductivity of the generated contexts. Via reverse generation, we augment the existing BAD dataset and construct a new dataset BAD+ which contains more than 120K diverse and highly inductive contexts in 12 categories. We test three popular pretrained dialogue models (Blender, DialoGPT and Plato2) and find that BAD+ can largely expose their safety problems. Furthermore, we show that BAD+ can greatly enhance the safety of generation, and we reveal the key factors of safety improvement. Our code and dataset is available at https://github.com/thu-coai/Reverse_Generation.

Among all the safety concerns that hinder the deployment of open-domain dialog systems (e.g., offensive languages, biases, and toxic behaviors), social bias presents an insidious challenge. Addressing this challenge requires rigorous analyses and normative reasoning. In this paper, we focus our investigation on social bias measurement to facilitate the development of unbiased dialog systems. We first propose a novel Dial-Bias Framework for analyzing the social bias in conversations using a holistic method beyond bias lexicons or dichotomous annotations. Leveraging the proposed framework, we further introduce the CDial-Bias Dataset which is, to the best of our knowledge, the first annotated Chinese social bias dialog dataset. We also establish a fine-grained dialog bias measurement benchmark and conduct in-depth ablation studies to shed light on the utility of the detailed annotations in the proposed dataset. Finally, we evaluate representative Chinese generative models with our classifiers to unveil the presence of social bias in these systems.

Dialogue safety problems severely limit the real-world deployment of neural conversational models and have attracted great research interests recently. However, dialogue safety problems remain under-defined and the corresponding dataset is scarce. We propose a taxonomy for dialogue safety specifically designed to capture unsafe behaviors in human-bot dialogue settings, with focuses on context-sensitive unsafety, which is under-explored in prior works. To spur research in this direction, we compile DiaSafety, a dataset with rich context-sensitive unsafe examples. Experiments show that existing safety guarding tools fail severely on our dataset. As a remedy, we train a dialogue safety classifier to provide a strong baseline for context-sensitive dialogue unsafety detection. With our classifier, we perform safety evaluations on popular conversational models and show that existing dialogue systems still exhibit concerning context-sensitive safety problems.

Offensive language detection is increasingly crucial for maintaining a civilized social media platform and deploying pre-trained language models. However, this task in Chinese is still under exploration due to the scarcity of reliable datasets. To this end, we propose a benchmark –COLD for Chinese offensive language analysis, including a Chinese Offensive Language Dataset –COLDATASET and a baseline detector –COLDETECTOR which is trained on the dataset. We show that the COLD benchmark contributes to Chinese offensive language detection which is challenging for existing resources. We then deploy the COLDETECTOR and conduct detailed analyses on popular Chinese pre-trained language models. We first analyze the offensiveness of existing generative models and show that these models inevitably expose varying degrees of offensive issues. Furthermore, we investigate the factors that influence the offensive generations, and we find that anti-bias contents and keywords referring to certain groups or revealing negative attitudes trigger offensive outputs easier.