MIRAGE-Bench: LLM Agent is Hallucinating and Where to Find Them

Hallucinations pose critical risks for large language model (LLM)-based agents, often manifesting as hallucinative actions resulting from fabricated or misinterpreted information within the cognitive context. While recent studies have exposed such failures, existing evaluations remain fragmented and lack a principled testbed. In this paper, we present MIRAGE-Bench--Measuring Illusions in Risky AGEnt settings--the first unified benchmark for eliciting and evaluating hallucinations in interactive LLM-agent scenarios. We begin by introducing a three-part taxonomy to address agentic hallucinations: actions that are unfaithful to (i) task instructions, (ii) execution history, or (iii) environment observations. To analyze, we first elicit such failures by performing a systematic audit of existing agent benchmarks, then synthesize test cases using a snapshot strategy that isolates decision points in deterministic and reproducible manners. To evaluate hallucination behaviors, we adopt a fine-grained-level LLM-as-a-Judge paradigm with tailored risk-aware prompts, enabling scalable, high-fidelity assessment of agent actions without enumerating full action spaces. MIRAGE-Bench provides actionable insights on failure modes of LLM agents and lays the groundwork for principled progress in mitigating hallucinations in interactive environments.

Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report

To understand and identify the unprecedented risks posed by rapidly advancing artificial intelligence (AI) models, this report presents a comprehensive assessment of their frontier risks. Drawing on the E-T-C analysis (deployment environment, threat source, enabling capability) from the Frontier AI Risk Management Framework (v1.0) (SafeWork-F1-Framework), we identify critical risks in seven areas: cyber offense, biological and chemical risks, persuasion and manipulation, uncontrolled autonomous AI R\&D, strategic deception and scheming, self-replication, and collusion. Guided by the "AI-$45^\circ$ Law," we evaluate these risks using "red lines" (intolerable thresholds) and "yellow lines" (early warning indicators) to define risk zones: green (manageable risk for routine deployment and continuous monitoring), yellow (requiring strengthened mitigations and controlled deployment), and red (necessitating suspension of development and/or deployment). Experimental results show that all recent frontier AI models reside in green and yellow zones, without crossing red lines. Specifically, no evaluated models cross the yellow line for cyber offense or uncontrolled AI R\&D risks. For self-replication, and strategic deception and scheming, most models remain in the green zone, except for certain reasoning models in the yellow zone. In persuasion and manipulation, most models are in the yellow zone due to their effective influence on humans. For biological and chemical risks, we are unable to rule out the possibility of most models residing in the yellow zone, although detailed threat modeling and in-depth assessment are required to make further claims. This work reflects our current understanding of AI frontier risks and urges collective action to mitigate these challenges.

Self-Paced Collaborative and Adversarial Network for Unsupervised Domain Adaptation

This paper proposes a new unsupervised domain adaptation approach called Collaborative and Adversarial Network (CAN), which uses the domain-collaborative and domain-adversarial learning strategy for training the neural network. The domain-collaborative learning aims to learn domain-specific feature representation to preserve the discriminability for the target domain, while the domain adversarial learning aims to learn domain-invariant feature representation to reduce the domain distribution mismatch between the source and target domains. We show that these two learning strategies can be uniformly formulated as domain classifier learning with positive or negative weights on the losses. We then design a collaborative and adversarial training scheme, which automatically learns domain-specific representations from lower blocks in CNNs through collaborative learning and domain-invariant representations from higher blocks through adversarial learning. Moreover, to further enhance the discriminability in the target domain, we propose Self-Paced CAN (SPCAN), which progressively selects pseudo-labeled target samples for re-training the classifiers. We employ a self-paced learning strategy to select pseudo-labeled target samples in an easy-to-hard fashion. Comprehensive experiments on different benchmark datasets, Office-31, ImageCLEF-DA, and VISDA-2017 for the object recognition task, and UCF101-10 and HMDB51-10 for the video action recognition task, show our newly proposed approaches achieve the state-of-the-art performance, which clearly demonstrates the effectiveness of our proposed approaches for unsupervised domain adaptation.

Progressive Modality Cooperation for Multi-Modality Domain Adaptation

In this work, we propose a new generic multi-modality domain adaptation framework called Progressive Modality Cooperation (PMC) to transfer the knowledge learned from the source domain to the target domain by exploiting multiple modality clues (\eg, RGB and depth) under the multi-modality domain adaptation (MMDA) and the more general multi-modality domain adaptation using privileged information (MMDA-PI) settings. Under the MMDA setting, the samples in both domains have all the modalities. In two newly proposed modules of our PMC, the multiple modalities are cooperated for selecting the reliable pseudo-labeled target samples, which captures the modality-specific information and modality-integrated information, respectively. Under the MMDA-PI setting, some modalities are missing in the target domain. Hence, to better exploit the multi-modality data in the source domain, we further propose the PMC with privileged information (PMC-PI) method by proposing a new multi-modality data generation (MMG) network. MMG generates the missing modalities in the target domain based on the source domain data by considering both domain distribution mismatch and semantics preservation, which are respectively achieved by using adversarial learning and conditioning on weighted pseudo semantics. Extensive experiments on three image datasets and eight video datasets for various multi-modality cross-domain visual recognition tasks under both MMDA and MMDA-PI settings clearly demonstrate the effectiveness of our proposed PMC framework.

CityNavAgent: Aerial Vision-and-Language Navigation with Hierarchical Semantic Planning and Global Memory

Aerial vision-and-language navigation (VLN), requiring drones to interpret natural language instructions and navigate complex urban environments, emerges as a critical embodied AI challenge that bridges human-robot interaction, 3D spatial reasoning, and real-world deployment. Although existing ground VLN agents achieved notable results in indoor and outdoor settings, they struggle in aerial VLN due to the absence of predefined navigation graphs and the exponentially expanding action space in long-horizon exploration. In this work, we propose \textbf{CityNavAgent}, a large language model (LLM)-empowered agent that significantly reduces the navigation complexity for urban aerial VLN. Specifically, we design a hierarchical semantic planning module (HSPM) that decomposes the long-horizon task into sub-goals with different semantic levels. The agent reaches the target progressively by achieving sub-goals with different capacities of the LLM. Additionally, a global memory module storing historical trajectories into a topological graph is developed to simplify navigation for visited targets. Extensive benchmark experiments show that our method achieves state-of-the-art performance with significant improvement. Further experiments demonstrate the effectiveness of different modules of CityNavAgent for aerial VLN in continuous city environments. The code is available at \href{https://github.com/VinceOuti/CityNavAgent}{link}.

The Point, the Vision and the Text: Does Point Cloud Boost Spatial Reasoning of Large Language Models?

3D Large Language Models (LLMs) leveraging spatial information in point clouds for 3D spatial reasoning attract great attention. Despite some promising results, the role of point clouds in 3D spatial reasoning remains under-explored. In this work, we comprehensively evaluate and analyze these models to answer the research question: \textit{Does point cloud truly boost the spatial reasoning capacities of 3D LLMs?} We first evaluate the spatial reasoning capacity of LLMs with different input modalities by replacing the point cloud with the visual and text counterparts. We then propose a novel 3D QA (Question-answering) benchmark, ScanReQA, that comprehensively evaluates models' understanding of binary spatial relationships. Our findings reveal several critical insights: 1) LLMs without point input could even achieve competitive performance even in a zero-shot manner; 2) existing 3D LLMs struggle to comprehend the binary spatial relationships; 3) 3D LLMs exhibit limitations in exploiting the structural coordinates in point clouds for fine-grained spatial reasoning. We think these conclusions can help the next step of 3D LLMs and also offer insights for foundation models in other modalities. We release datasets and reproducible codes in the anonymous project page: https://3d-llm.xyz.

UrbanVideo-Bench: Benchmarking Vision-Language Models on Embodied Intelligence with Video Data in Urban Spaces

Large multimodal models exhibit remarkable intelligence, yet their embodied cognitive abilities during motion in open-ended urban 3D space remain to be explored. We introduce a benchmark to evaluate whether video-large language models (Video-LLMs) can naturally process continuous first-person visual observations like humans, enabling recall, perception, reasoning, and navigation. We have manually control drones to collect 3D embodied motion video data from real-world cities and simulated environments, resulting in 1.5k video clips. Then we design a pipeline to generate 5.2k multiple-choice questions. Evaluations of 17 widely-used Video-LLMs reveal current limitations in urban embodied cognition. Correlation analysis provides insight into the relationships between different tasks, showing that causal reasoning has a strong correlation with recall, perception, and navigation, while the abilities for counterfactual and associative reasoning exhibit lower correlation with other tasks. We also validate the potential for Sim-to-Real transfer in urban embodiment through fine-tuning.

Understanding and Evaluating Hallucinations in 3D Visual Language Models

Recently, 3D-LLMs, which combine point-cloud encoders with large models, have been proposed to tackle complex tasks in embodied intelligence and scene understanding. In addition to showing promising results on 3D tasks, we found that they are significantly affected by hallucinations. For instance, they may generate objects that do not exist in the scene or produce incorrect relationships between objects. To investigate this issue, this work presents the first systematic study of hallucinations in 3D-LLMs. We begin by quickly evaluating hallucinations in several representative 3D-LLMs and reveal that they are all significantly affected by hallucinations. We then define hallucinations in 3D scenes and, through a detailed analysis of datasets, uncover the underlying causes of these hallucinations. We find three main causes: (1) Uneven frequency distribution of objects in the dataset. (2) Strong correlations between objects. (3) Limited diversity in object attributes. Additionally, we propose new evaluation metrics for hallucinations, including Random Point Cloud Pair and Opposite Question Evaluations, to assess whether the model generates responses based on visual information and aligns it with the text's meaning.

EmbodiedCity: A Benchmark Platform for Embodied Agent in Real-world City Environment

Embodied artificial intelligence emphasizes the role of an agent's body in generating human-like behaviors. The recent efforts on EmbodiedAI pay a lot of attention to building up machine learning models to possess perceiving, planning, and acting abilities, thereby enabling real-time interaction with the world. However, most works focus on bounded indoor environments, such as navigation in a room or manipulating a device, with limited exploration of embodying the agents in open-world scenarios. That is, embodied intelligence in the open and outdoor environment is less explored, for which one potential reason is the lack of high-quality simulators, benchmarks, and datasets. To address it, in this paper, we construct a benchmark platform for embodied intelligence evaluation in real-world city environments. Specifically, we first construct a highly realistic 3D simulation environment based on the real buildings, roads, and other elements in a real city. In this environment, we combine historically collected data and simulation algorithms to conduct simulations of pedestrian and vehicle flows with high fidelity. Further, we designed a set of evaluation tasks covering different EmbodiedAI abilities. Moreover, we provide a complete set of input and output interfaces for access, enabling embodied agents to easily take task requirements and current environmental observations as input and then make decisions and obtain performance evaluations. On the one hand, it expands the capability of existing embodied intelligence to higher levels. On the other hand, it has a higher practical value in the real world and can support more potential applications for artificial general intelligence. Based on this platform, we evaluate some popular large language models for embodied intelligence capabilities of different dimensions and difficulties.

GUI Action Narrator: Where and When Did That Action Take Place?

The advent of Multimodal LLMs has significantly enhanced image OCR recognition capabilities, making GUI automation a viable reality for increasing efficiency in digital tasks. One fundamental aspect of developing a GUI automation system is understanding primitive GUI actions. This comprehension is crucial as it enables agents to learn from user demonstrations, an essential element of automation. To rigorously evaluate such capabilities, we developed a video captioning benchmark for GUI actions, comprising 4,189 diverse video captioning samples. This task presents unique challenges compared to natural scene video captioning: 1) GUI screenshots typically contain denser information than natural scenes, and 2) events within GUIs are subtler and occur more rapidly, requiring precise attention to the appropriate time span and spatial region for accurate understanding. To address these challenges, we introduce our GUI action dataset \textbf{Act2Cap} as well as a simple yet effective framework, \textbf{GUI Narrator}, for GUI video captioning that utilizes the cursor as a visual prompt to enhance the interpretation of high-resolution screenshots. Specifically, a cursor detector is trained on our dataset, and a multimodal LLM model with mechanisms for selecting keyframes and key regions generates the captions. Experimental results indicate that even for today's most advanced multimodal models, such as GPT-4o, the task remains highly challenging. Additionally, our evaluations show that our strategy effectively enhances model performance, whether integrated into the fine-tuning of open-source models or employed as a prompting strategy in closed-source models.