Human processes video reasoning in a sequential spatio-temporal reasoning logic, we first identify the relevant frames ("when") and then analyse the spatial relationships ("where") between key objects, and finally leverage these relationships to draw inferences ("what"). However, can Video Large Language Models (Video-LLMs) also "reason through a sequential spatio-temporal logic" in videos? Existing Video-LLM benchmarks primarily focus on assessing object presence, neglecting relational reasoning. Consequently, it is difficult to measure whether a model truly comprehends object interactions (actions/events) in videos or merely relies on pre-trained "memory" of co-occurrences as biases in generating answers. In this work, we introduce a Video Spatio-Temporal Reasoning (V-STaR) benchmark to address these shortcomings. The key idea is to decompose video understanding into a Reverse Spatio-Temporal Reasoning (RSTR) task that simultaneously evaluates what objects are present, when events occur, and where they are located while capturing the underlying Chain-of-thought (CoT) logic. To support this evaluation, we construct a dataset to elicit the spatial-temporal reasoning process of Video-LLMs. It contains coarse-to-fine CoT questions generated by a semi-automated GPT-4-powered pipeline, embedding explicit reasoning chains to mimic human cognition. Experiments from 14 Video-LLMs on our V-STaR reveal significant gaps between current Video-LLMs and the needs for robust and consistent spatio-temporal reasoning.
Plug-and-play diffusion priors (PnPDP) have emerged as a promising research direction for solving inverse problems. However, current studies primarily focus on natural image restoration, leaving the performance of these algorithms in scientific inverse problems largely unexplored. To address this gap, we introduce \textsc{InverseBench}, a framework that evaluates diffusion models across five distinct scientific inverse problems. These problems present unique structural challenges that differ from existing benchmarks, arising from critical scientific applications such as optical tomography, medical imaging, black hole imaging, seismology, and fluid dynamics. With \textsc{InverseBench}, we benchmark 14 inverse problem algorithms that use plug-and-play diffusion priors against strong, domain-specific baselines, offering valuable new insights into the strengths and weaknesses of existing algorithms. To facilitate further research and development, we open-source the codebase, along with datasets and pre-trained models, at https://devzhk.github.io/InverseBench/.
Reinforcement Learning (RL) has proven largely effective in obtaining stable locomotion gaits for legged robots. However, designing control algorithms which can robustly navigate unseen environments with obstacles remains an ongoing problem within quadruped locomotion. To tackle this, it is convenient to solve navigation tasks by means of a hierarchical approach with a low-level locomotion policy and a high-level navigation policy. Crucially, the high-level policy needs to be robust to dynamic obstacles along the path of the agent. In this work, we propose a novel way to endow navigation policies with robustness by a training process that models obstacles as adversarial agents, following the adversarial RL paradigm. Importantly, to improve the reliability of the training process, we bound the rationality of the adversarial agent resorting to quantal response equilibria, and place a curriculum over its rationality. We called this method Hierarchical policies via Quantal response Adversarial Reinforcement Learning (Hi-QARL). We demonstrate the robustness of our method by benchmarking it in unseen randomized mazes with multiple obstacles. To prove its applicability in real scenarios, our method is applied on a Unitree GO1 robot in simulation.
The ability of robots to recognize human gestures facilitates a natural and accessible human-robot collaboration. However, most work in gesture recognition remains rooted in reference frame-dependent representations. This poses a challenge when reference frames vary due to different work cell layouts, imprecise frame calibrations, or other environmental changes. This paper investigated the use of invariant trajectory descriptors for robust hand palm motion gesture recognition under reference frame changes. First, a novel dataset of recorded Hand Palm Motion (HPM) gestures is introduced. The motion gestures in this dataset were specifically designed to be distinguishable without dependence on specific reference frames or directional cues. Afterwards, multiple invariant trajectory descriptor approaches were benchmarked to assess how their performances generalize to this novel HPM dataset. After this offline benchmarking, the best scoring approach is validated for online recognition by developing a real-time Proof of Concept (PoC). In this PoC, hand palm motion gestures were used to control the real-time movement of a manipulator arm. The PoC demonstrated a high recognition reliability in real-time operation, achieving an $F_1$-score of 92.3%. This work demonstrates the effectiveness of the invariant descriptor approach as a standalone solution. Moreover, we believe that the invariant descriptor approach can also be utilized within other state-of-the-art pattern recognition and learning systems to improve their robustness against reference frame variations.
Agentic workflows invoked by Large Language Models (LLMs) have achieved remarkable success in handling complex tasks. However, optimizing such workflows is costly and inefficient in real-world applications due to extensive invocations of LLMs. To fill this gap, this position paper formulates agentic workflows as computational graphs and advocates Graph Neural Networks (GNNs) as efficient predictors of agentic workflow performances, avoiding repeated LLM invocations for evaluation. To empirically ground this position, we construct FLORA-Bench, a unified platform for benchmarking GNNs for predicting agentic workflow performances. With extensive experiments, we arrive at the following conclusion: GNNs are simple yet effective predictors. This conclusion supports new applications of GNNs and a novel direction towards automating agentic workflow optimization. All codes, models, and data are available at https://github.com/youngsoul0731/Flora-Bench.
We present a benchmarking study of vision-based robotic grasping algorithms with distinct approaches, and provide a comparative analysis. In particular, we compare two machine-learning-based and two analytical algorithms using an existing benchmarking protocol from the literature and determine the algorithm's strengths and weaknesses under different experimental conditions. These conditions include variations in lighting, background textures, cameras with different noise levels, and grippers. We also run analogous experiments in simulations and with real robots and present the discrepancies. Some experiments are also run in two different laboratories using same protocols to further analyze the repeatability of our results. We believe that this study, comprising 5040 experiments, provides important insights into the role and challenges of systematic experimentation in robotic manipulation, and guides the development of new algorithms by considering the factors that could impact the performance. The experiment recordings and our benchmarking software are publicly available.
We introduce DarkBench, a comprehensive benchmark for detecting dark design patterns--manipulative techniques that influence user behavior--in interactions with large language models (LLMs). Our benchmark comprises 660 prompts across six categories: brand bias, user retention, sycophancy, anthropomorphism, harmful generation, and sneaking. We evaluate models from five leading companies (OpenAI, Anthropic, Meta, Mistral, Google) and find that some LLMs are explicitly designed to favor their developers' products and exhibit untruthful communication, among other manipulative behaviors. Companies developing LLMs should recognize and mitigate the impact of dark design patterns to promote more ethical AI.
In this paper, we propose a comprehensive evaluation benchmark for Visual Language Models (VLM) in Traditional Chinese. Our evaluation suite, the first of its kind, contains two complementary components: (1) VisTai-MCQ, a collection of manually curated exam multi-choice questions from 21 academic subjects designed to test the broad knowledge and reasoning capabilities of VLMs; and (2) VisTai-Dialogue, an open dialogue benchmark comprising 131 image-question pairs manually created to evaluate VLMs' ability in free-form dialogue generation within Taiwanese cultural contexts. These benchmarks address a critical gap in the evaluation landscape, where existing benchmarks predominantly focus on English or Simplified Chinese, neglecting the unique linguistic and cultural aspects of Traditional Chinese used in regions like Taiwan and Hong Kong. Our analysis reveals significant performance differences across various VLMs and highlights specific challenges in processing Traditional Chinese visual content.
Video large language models have achieved remarkable performance in tasks such as video question answering, however, their temporal understanding remains suboptimal. To address this limitation, we curate a dedicated instruction fine-tuning dataset that focuses on enhancing temporal comprehension across five key dimensions. In order to reduce reliance on costly temporal annotations, we introduce a multi-task prompt fine-tuning approach that seamlessly integrates temporal-sensitive tasks into existing instruction datasets without requiring additional annotations. Furthermore, we develop a novel benchmark for temporal-sensitive video understanding that not only fills the gaps in dimension coverage left by existing benchmarks but also rigorously filters out potential shortcuts, ensuring a more accurate evaluation. Extensive experimental results demonstrate that our approach significantly enhances the temporal understanding of video-LLMs while avoiding reliance on shortcuts.
Large Multimodal Models (LMMs) are increasingly vulnerable to AI-generated extremist content, including photorealistic images and text, which can be used to bypass safety mechanisms and generate harmful outputs. However, existing datasets for evaluating LMM robustness offer limited exploration of extremist content, often lacking AI-generated images, diverse image generation models, and comprehensive coverage of historical events, which hinders a complete assessment of model vulnerabilities. To fill this gap, we introduce ExtremeAIGC, a benchmark dataset and evaluation framework designed to assess LMM vulnerabilities against such content. ExtremeAIGC simulates real-world events and malicious use cases by curating diverse text- and image-based examples crafted using state-of-the-art image generation techniques. Our study reveals alarming weaknesses in LMMs, demonstrating that even cutting-edge safety measures fail to prevent the generation of extremist material. We systematically quantify the success rates of various attack strategies, exposing critical gaps in current defenses and emphasizing the need for more robust mitigation strategies.
Accurate and efficient 3D segmentation is essential for both clinical and research applications. While foundation models like SAM have revolutionized interactive segmentation, their 2D design and domain shift limitations make them ill-suited for 3D medical images. Current adaptations address some of these challenges but remain limited, either lacking volumetric awareness, offering restricted interactivity, or supporting only a small set of structures and modalities. Usability also remains a challenge, as current tools are rarely integrated into established imaging platforms and often rely on cumbersome web-based interfaces with restricted functionality. We introduce nnInteractive, the first comprehensive 3D interactive open-set segmentation method. It supports diverse prompts-including points, scribbles, boxes, and a novel lasso prompt-while leveraging intuitive 2D interactions to generate full 3D segmentations. Trained on 120+ diverse volumetric 3D datasets (CT, MRI, PET, 3D Microscopy, etc.), nnInteractive sets a new state-of-the-art in accuracy, adaptability, and usability. Crucially, it is the first method integrated into widely used image viewers (e.g., Napari, MITK), ensuring broad accessibility for real-world clinical and research applications. Extensive benchmarking demonstrates that nnInteractive far surpasses existing methods, setting a new standard for AI-driven interactive 3D segmentation. nnInteractive is publicly available: https://github.com/MIC-DKFZ/napari-nninteractive (Napari plugin), https://www.mitk.org/MITK-nnInteractive (MITK integration), https://github.com/MIC-DKFZ/nnInteractive (Python backend).
Recent image generation schemes typically capture image distribution in a pre-constructed latent space relying on a frozen image tokenizer. Though the performance of tokenizer plays an essential role to the successful generation, its current evaluation metrics (e.g. rFID) fail to precisely assess the tokenizer and correlate its performance to the generation quality (e.g. gFID). In this paper, we comprehensively analyze the reason for the discrepancy of reconstruction and generation qualities in a discrete latent space, and, from which, we propose a novel plug-and-play tokenizer training scheme to facilitate latent space construction. Specifically, a latent perturbation approach is proposed to simulate sampling noises, i.e., the unexpected tokens sampled, from the generative process. With the latent perturbation, we further propose (1) a novel tokenizer evaluation metric, i.e., pFID, which successfully correlates the tokenizer performance to generation quality and (2) a plug-and-play tokenizer training scheme, which significantly enhances the robustness of tokenizer thus boosting the generation quality and convergence speed. Extensive benchmarking are conducted with 11 advanced discrete image tokenizers with 2 autoregressive generation models to validate our approach. The tokenizer trained with our proposed latent perturbation achieve a notable 1.60 gFID with classifier-free guidance (CFG) and 3.45 gFID without CFG with a $\sim$400M generator. Code: https://github.com/lxa9867/ImageFolder.
Predictive modelling is vital to guide preventive efforts. Whilst large-scale prospective cohort studies and a diverse toolkit of available machine learning (ML) algorithms have facilitated such survival task efforts, choosing the best-performing algorithm remains challenging. Benchmarking studies to date focus on relatively small-scale datasets and it is unclear how well such findings translate to large datasets that combine omics and clinical features. We sought to benchmark eight distinct survival task implementations, ranging from linear to deep learning (DL) models, within the large-scale prospective cohort study UK Biobank (UKB). We compared discrimination and computational requirements across heterogenous predictor matrices and endpoints. Finally, we assessed how well different architectures scale with sample sizes ranging from n = 5,000 to n = 250,000 individuals. Our results show that discriminative performance across a multitude of metrices is dependent on endpoint frequency and predictor matrix properties, with very robust performance of (penalised) COX Proportional Hazards (COX-PH) models. Of note, there are certain scenarios which favour more complex frameworks, specifically if working with larger numbers of observations and relatively simple predictor matrices. The observed computational requirements were vastly different, and we provide solutions in cases where current implementations were impracticable. In conclusion, this work delineates how optimal model choice is dependent on a variety of factors, including sample size, endpoint frequency and predictor matrix properties, thus constituting an informative resource for researchers working on similar datasets. Furthermore, we showcase how linear models still display a highly effective and scalable platform to perform risk modelling at scale and suggest that those are reported alongside non-linear ML models.
This work presents the MarineGym, a high-performance reinforcement learning (RL) platform specifically designed for underwater robotics. It aims to address the limitations of existing underwater simulation environments in terms of RL compatibility, training efficiency, and standardized benchmarking. MarineGym integrates a proposed GPU-accelerated hydrodynamic plugin based on Isaac Sim, achieving a rollout speed of 250,000 frames per second on a single NVIDIA RTX 3060 GPU. It also provides five models of unmanned underwater vehicles (UUVs), multiple propulsion systems, and a set of predefined tasks covering core underwater control challenges. Additionally, the DR toolkit allows flexible adjustments of simulation and task parameters during training to improve Sim2Real transfer. Further benchmark experiments demonstrate that MarineGym improves training efficiency over existing platforms and supports robust policy adaptation under various perturbations. We expect this platform could drive further advancements in RL research for underwater robotics. For more details about MarineGym and its applications, please visit our project page: https://marine-gym.com/.
This paper presents Ev-Layout, a novel large-scale event-based multi-modal dataset designed for indoor layout estimation and tracking. Ev-Layout makes key contributions to the community by: Utilizing a hybrid data collection platform (with a head-mounted display and VR interface) that integrates both RGB and bio-inspired event cameras to capture indoor layouts in motion. Incorporating time-series data from inertial measurement units (IMUs) and ambient lighting conditions recorded during data collection to highlight the potential impact of motion speed and lighting on layout estimation accuracy. The dataset consists of 2.5K sequences, including over 771.3K RGB images and 10 billion event data points. Of these, 39K images are annotated with indoor layouts, enabling research in both event-based and video-based indoor layout estimation. Based on the dataset, we propose an event-based layout estimation pipeline with a novel event-temporal distribution feature module to effectively aggregate the spatio-temporal information from events. Additionally, we introduce a spatio-temporal feature fusion module that can be easily integrated into a transformer module for fusion purposes. Finally, we conduct benchmarking and extensive experiments on the Ev-Layout dataset, demonstrating that our approach significantly improves the accuracy of dynamic indoor layout estimation compared to existing event-based methods.
3D point cloud classification, a task in computer vision, has recently gained substantial interest due to its extensive applications across various domains such as augmented and virtual reality, robotics, and autonomous driving. However, classifying 3D point clouds becomes challenging in real-world scenarios where the density levels vary. These variations are due to limitations inherent to current sensing technologies and variable environmental conditions, which often prevent the capture of detailed and comprehensive geometric representations. Existing classification methods, which have demonstrated impressive results on dense point clouds, may be affected and could experience reduced performance when dealing with low-density point clouds. Addressing this challenge, this paper introduces LD-PointNet++, an architecture optimized for the classification of low-density point clouds. To pave the way for LD-PointNet++, we conduct the first benchmarking study that systematically evaluates state-of-the-art point cloud classification methods across varying densities, revealing significant performance declines in low-density environments. Experimental results on the ModelNet benchmark dataset demonstrate that LD-PointNet++ not only outperforms these methods in low-density scenarios but also demonstrates consistent accuracy above 90 % across different density levels.
Identifying vulnerabilities in source code is crucial, especially in critical software components. Existing methods such as static analysis, dynamic analysis, formal verification, and recently Large Language Models are widely used to detect security flaws. This paper introduces CASTLE (CWE Automated Security Testing and Low-Level Evaluation), a benchmarking framework for evaluating the vulnerability detection capabilities of different methods. We assess 13 static analysis tools, 10 LLMs, and 2 formal verification tools using a hand-crafted dataset of 250 micro-benchmark programs covering 25 common CWEs. We propose the CASTLE Score, a novel evaluation metric to ensure fair comparison. Our results reveal key differences: ESBMC (a formal verification tool) minimizes false positives but struggles with vulnerabilities beyond model checking, such as weak cryptography or SQL injection. Static analyzers suffer from high false positives, increasing manual validation efforts for developers. LLMs perform exceptionally well in the CASTLE dataset when identifying vulnerabilities in small code snippets. However, their accuracy declines, and hallucinations increase as the code size grows. These results suggest that LLMs could play a pivotal role in future security solutions, particularly within code completion frameworks, where they can provide real-time guidance to prevent vulnerabilities. The dataset is accessible at https://github.com/CASTLE-Benchmark.
As Large Language Models (LLMs) become increasingly powerful and accessible to human users, ensuring fairness across diverse demographic groups, i.e., group fairness, is a critical ethical concern. However, current fairness and bias research in LLMs is limited in two aspects. First, compared to traditional group fairness in machine learning classification, it requires that the non-sensitive attributes, in this case, the prompt questions, be the same across different groups. In many practical scenarios, different groups, however, may prefer different prompt questions and this requirement becomes impractical. Second, it evaluates group fairness only for the LLM's final output without identifying the source of possible bias. Namely, the bias in LLM's output can result from both the pretraining and the finetuning. For finetuning, the bias can result from both the RLHF procedure and the learned reward model. Arguably, evaluating the group fairness of each component in the LLM pipeline could help develop better methods to mitigate the possible bias. Recognizing those two limitations, this work benchmarks the group fairness of learned reward models. By using expert-written text from arXiv, we are able to benchmark the group fairness of reward models without requiring the same prompt questions across different demographic groups. Surprisingly, our results demonstrate that all the evaluated reward models (e.g., Nemotron-4-340B-Reward, ArmoRM-Llama3-8B-v0.1, and GRM-llama3-8B-sftreg) exhibit statistically significant group unfairness. We also observed that top-performing reward models (w.r.t. canonical performance metrics) tend to demonstrate better group fairness.
Keeping large language models factually up-to-date is crucial for deployment, yet costly retraining remains a challenge. Knowledge editing offers a promising alternative, but methods are only tested on small-scale or synthetic edit benchmarks. In this work, we aim to bridge research into lifelong knowledge editing to real-world edits at practically relevant scale. We first introduce WikiBigEdit; a large-scale benchmark of real-world Wikidata edits, built to automatically extend lifelong for future-proof benchmarking. In its first instance, it includes over 500K question-answer pairs for knowledge editing alongside a comprehensive evaluation pipeline. Finally, we use WikiBigEdit to study existing knowledge editing techniques' ability to incorporate large volumes of real-world facts and contrast their capabilities to generic modification techniques such as retrieval augmentation and continual finetuning to acquire a complete picture of the practical extent of current lifelong knowledge editing.
One-class anomaly detection aims to detect objects that do not belong to a predefined normal class. In practice training data lack those anomalous samples; hence state-of-the-art methods are trained to discriminate between normal and synthetically-generated pseudo-anomalous data. Most methods use data augmentation techniques on normal images to simulate anomalies. However the best-performing ones implicitly leverage a geometric bias present in the benchmarking datasets. This limits their usability in more general conditions. Others are relying on basic noising schemes that may be suboptimal in capturing the underlying structure of normal data. In addition most still favour the image domain to generate pseudo-anomalies training models end-to-end from only the normal class and overlooking richer representations of the information. To overcome these limitations we consider frozen yet rich feature spaces given by pretrained models and create pseudo-anomalous features with a novel adaptive linear feature perturbation technique. It adapts the noise distribution to each sample applies decaying linear perturbations to feature vectors and further guides the classification process using a contrastive learning objective. Experimental evaluation conducted on both standard and geometric bias-free datasets demonstrates the superiority of our approach with respect to comparable baselines. The codebase is accessible via our public repository.