arXiv subjects

Andreas Bulling

Publications and source records attributed to Andreas Bulling.

At least 19 recordsLinked to original sources

arXiv2026

Unsupervised Partner Design Enables Robust Ad-hoc Teamwork

We introduce Unsupervised Partner Design (UPD), a population-free multi-agent reinforcement learning method for robust ad-hoc teamwork. UPD generates training partners on-the-fly and selects them adaptively based on a learnability criterion, removing the need for pre-trained partner populations or manual parameter tuning. We show that this simple mechanism enables effective partner diversity and can be extended to joint partner-environment selection when a procedural level generator is available. Across Level-Based Foraging, Overcooked-AI, and the Overcooked Generalisation Challenge, UPD consistently achieves strong performance compared to both population-based and population-free baselines. In a human-AI user study, agents trained with UPD achieve higher returns and are rated as more adaptive, more human-like, and less frustrating than all evaluated baseline methods.

cs.LG↗

arXiv2026

CARD: Diagnosing Belief to Action Routing Failures in Vision Language Models

Linear probes and activation steering have uncovered that vision-language models (VLMs) internally represent mental states such as agents' beliefs, knowledge, and intentions. However, it is unclear whether and how these representations are used by downstream predictions along these axes. To close this gap, we introduce Cross-Axis Routing Diagnostic (CARD), which steers activations along one axis while measuring the response of a different axis's prediction. Applied to open-weight VLMs on Relay Chain -- a new cooperative grid-world benchmark we propose -- we diagnose a critical routing failure: models fail to incorporate belief representations into their next action prediction, effectively leaving valuable information about their partners unused.

cs.CV↗

arXiv2026

The Yokai Learning Environment: Tracking Beliefs Over Space and Time

The ability to cooperate with unknown partners is a central challenge in cooperative AI and widely studied in the form of zero-shot coordination (ZSC), which evaluates an algorithm by measuring the performance of independently trained agents when paired. The Hanabi Learning Environment (HLE) has become the dominant benchmark for ZSC, but recent work has achieved near-perfect inter-seed cross-play performance, limiting its ability to track algorithmic progress. We introduce the Yokai Learning Environment (YLE) - an open-source multi-agent RL benchmark in which effective collaboration requires building common ground by tracking and updating beliefs over moving cards, reasoning under ambiguous hints, and deciding when to terminate the game based on inferred shared knowledge - features absent in the HLE, where beliefs are tied to hand slots and hints are truthful by rule. We evaluate the leading ZSC methods, including High-Entropy IPPO, Other-Play, and Off-Belief Learning, which achieve near-perfect inter-seed cross-play in the HLE, and show that in the YLE they exhibit persistent SP-XP gaps, degraded early-ending calibration, and weaker belief representations in cross-play, indicating failure to maintain consistent internal models with unseen partners. Methods that perform best in the HLE do not perform best in the YLE, indicating that progress measured on a single benchmark may not generalise. Together, these results establish YLE as a challenging new ZSC benchmark.

cs.AI↗

arXiv2026

MEAL: A Benchmark for Continual Multi-Agent Reinforcement Learning

Benchmarks play a central role in reinforcement learning (RL) research, yet their computational constraints often shape what is studied. Despite the motivation of lifelong learning, most continual RL papers consider only 3-10 sequential tasks, as CPU-bound environments make longer sequences impractical. Meanwhile, continual learning in cooperative multi-agent settings remains largely unexplored. To address these gaps, we introduce MEAL (Multi-agent Environments for Adaptive Learning), the first benchmark for continual multi-agent RL. By leveraging JAX and GPU acceleration, MEAL enables training on sequences of 100 tasks in a few hours on a single GPU. We find that long task sequences reveal failure modes that do not appear at smaller scales.

cs.AI↗

arXiv2026

CLAD: Constrained Latent Action Diffusion for Vision-Language Procedure Planning

We propose CLAD, a Constrained Latent Action Diffusion model for vision-language procedure planning in instructional videos. Procedure planning is the challenging task of predicting intermediate actions given a visual observation of a start and a goal state. However, future interactive AI systems must also be able to plan procedures using multi-modal input, e.g., where visual observations are augmented with language descriptions. To tackle this vision-language procedure planning task, our method uses a Variational Autoencoder (VAE) to learn the latent representation of actions and observations as constraints and integrate them into the diffusion process. This approach exploits that the latent space of diffusion models already has semantics that can be used. We use the latent constraints to steer the diffusion model to better generate actions. We report extensive experiments on the popular CrossTask, Coin, and NIV datasets and show that our method outperforms state-of-the-art methods by a large margin. By evaluating ablated versions of our method, we further show that the proposed integration of the action and observation representations learnt in the VAE latent space is key to these performance improvements.

cs.CV↗

arXiv2026

High entropy leads to symmetry-equivariant policies in Dec-POMDPs

We prove that in any Dec-POMDP, sufficiently high entropy regularization ensures that the policy gradient flow with tabular softmax parametrization always converges, for any initialization, to the same joint policy, and that this joint policy is equivariant w.r.t. all symmetries of the Dec-POMDP. In particular, policies coming from different initializations will be fully compatible, in that their cross-play returns are equal to their self-play returns. Through extensive evaluation of independent PPO, arguably the standard baseline deep multi-agent policy gradient algorithm, in the Hanabi, Overcooked and Yokai environments, we find that the entropy coefficient has a massive influence on the cross-play returns between independently trained policies, and that the decrease in self-play returns coming from increased entropy regularization can often be counteracted by greedifying the learned policies after training. In Hanabi in particular we achieve a new SOTA in inter-seed cross-play this way. While we give examples of Dec-POMDPs in which one cannot learn the optimal symmetry-equivariant policy this way, both our theoretical and empirical results suggest that one should consider far higher entropy coefficients during hyperparameter sweeps in Dec-POMDPs than is typically done. Code for our experiments can be found at https://github.com/jforkel/JAX-OBL

cs.LG↗

arXiv2026

RobustSpring: Benchmarking Robustness to Image Corruptions for Optical Flow, Scene Flow and Stereo

Standard benchmarks for optical flow, scene flow, and stereo vision algorithms generally focus on model accuracy rather than robustness to image corruptions like noise or rain. Hence, the resilience of models to such real-world perturbations is largely unquantified. To address this, we present RobustSpring, a comprehensive dataset and benchmark for evaluating robustness to image corruptions for optical flow, scene flow, and stereo models. RobustSpring applies 20 different image corruptions, including noise, blur, color changes, quality degradations, and weather distortions, in a time-, stereo-, and depth-consistent manner to the high-resolution Spring dataset, creating a suite of 20,000 corrupted images that reflect challenging conditions. RobustSpring enables comparisons of model robustness via a new corruption robustness metric. Integration with the Spring benchmark enables two-axis evaluations of both accuracy and robustness. We benchmark a curated selection of initial models, observing that robustness varies widely by corruption type, and experimentally show that evaluations on RobustSpring indicate real-world robustness. RobustSpring is a new computer vision benchmark to treat robustness as a first-class citizen, fostering models that are accurate and resilient. It is available at https://spring-benchmark.org.

cs.CV↗

arXiv2026

Ontology-Guided Diffusion for Zero-Shot Visual Sim2Real Transfer

Bridging the simulation-to-reality (sim2real) gap remains challenging as labelled real-world data is scarce. Existing diffusion-based approaches rely on unstructured prompts or statistical alignment, which do not capture the structured factors that make images look real. We introduce Ontology- Guided Diffusion (OGD), a neuro-symbolic zero-shot sim2real image translation framework that represents realism as structured knowledge. OGD decomposes realism into an ontology of interpretable traits -- such as lighting and material properties -- and encodes their relationships in a knowledge graph. From a synthetic image, OGD infers trait activations and uses a graph neural network to produce a global embedding. In parallel, a symbolic planner uses the ontology traits to compute a consistent sequence of visual edits needed to narrow the realism gap. The graph embedding conditions a pretrained instruction-guided diffusion model via cross-attention, while the planned edits are converted into a structured instruction prompt. Across benchmarks, our graph-based embeddings better distinguish real from synthetic imagery than baselines, and OGD outperforms state-of-the-art diffusion methods in sim2real image translations. Overall, OGD shows that explicitly encoding realism structure enables interpretable, data-efficient, and generalisable zero-shot sim2real transfer.

cs.CV↗

arXiv2026

QualitEye: Public and Privacy-preserving Gaze Data Quality Verification

Gaze-based applications are increasingly advancing with the availability of large datasets but ensuring data quality presents a substantial challenge when collecting data at scale. It further requires different parties to collaborate, therefore, privacy concerns arise. We propose QualitEye--the first method for verifying image-based gaze data quality. QualitEye employs a new semantic representation of eye images that contains the information required for verification while excluding irrelevant information for better domain adaptation. QualitEye covers a public setting where parties can freely exchange data and a privacy-preserving setting where parties cannot reveal their raw data nor derive gaze features/labels of others with adapted private set intersection protocols. We evaluate QualitEye on the MPIIFaceGaze and GazeCapture datasets and achieve a high verification performance (with a small overhead in runtime for privacy-preserving versions). Hence, QualitEye paves the way for new gaze analysis methods at the intersection of machine learning, human-computer interaction, and cryptography.

cs.HC↗

arXiv2026

ObjectVisA-120: Object-based Visual Attention Prediction in Interactive Street-crossing Environments

The object-based nature of human visual attention is well-known in cognitive science, but has only played a minor role in computational visual attention models so far. This is mainly due to a lack of suitable datasets and evaluation metrics for object-based attention. To address these limitations, we present ObjectVisA-120 -- a novel 120-participant dataset of spatial street-crossing navigation in virtual reality specifically geared to object-based attention evaluations. The uniqueness of the presented dataset lies in the ethical and safety affiliated challenges that make collecting comparable data in real-world environments highly difficult. ObjectVisA-120 not only features accurate gaze data and a complete state-space representation of objects in the virtual environment, but it also offers variable scenario complexities and rich annotations, including panoptic segmentation, depth information, and vehicle keypoints. We further propose object-based similarity (oSIM) as a novel metric to evaluate the performance of object-based visual attention models, a previously unexplored performance characteristic. Our evaluations show that explicitly optimising for object-based attention not only improves oSIM performance but also leads to an improved model performance on common metrics. In addition, we present SUMGraph, a Mamba U-Net-based model, which explicitly encodes critical scene objects (vehicles) in a graph representation, leading to further performance improvements over several state-of-the-art visual attention prediction methods. The dataset, code and models will be publicly released.

cs.CV↗

arXiv2026

ChartOptimiser: Task-driven Optimisation of Chart Designs

Automated chart design has seen significant advancements with the emergence of Large-Language Models (LLMs), which offer a practical solution for generating charts. However, LLMs frequently introduce possibly critical design failures, such as data manipulation and confabulation. While expert users can potentially mitigate these issues through iterative prompt engineering, this process requires substantial design knowledge and significant effort, remaining a massive barrier for the general public. In this paper, we present ChartOptimiser, an automated method for generating chart designs with fidelity, efficiency, and effectiveness. Given the inter-dependencies between individual design parameters, ChartOptimiser employs Bayesian optimisation to effectively search the chart design space for a novel objective function grounded in four perceptual metrics. Our empirical evaluations in bar and pie charts demonstrate that ChartOptimiser eliminates iterative design loops, providing non-expert users with high-quality charts that outperform LLM-generated designs in chart clarity, task-solving ease, and visual aesthetics.

cs.HC↗

arXiv2025

CryptoQA: A Large-scale Question-answering Dataset for AI-assisted Cryptography

Large language models (LLMs) excel at many general-purpose natural language processing tasks. However, their ability to perform deep reasoning and mathematical analysis, particularly for complex tasks as required in cryptography, remains poorly understood, largely due to the lack of suitable data for evaluation and training. To address this gap, we present CryptoQA, the first large-scale question-answering (QA) dataset specifically designed for cryptography. CryptoQA contains over two million QA pairs drawn from curated academic sources, along with contextual metadata that can be used to test the cryptographic capabilities of LLMs and to train new LLMs on cryptographic tasks. We benchmark 15 state-of-the-art LLMs on CryptoQA, evaluating their factual accuracy, mathematical reasoning, consistency, referencing, backward reasoning, and robustness to adversarial samples. In addition to quantitative metrics, we provide expert reviews that qualitatively assess model outputs and establish a gold-standard baseline. Our results reveal significant performance deficits of LLMs, particularly on tasks that require formal reasoning and precise mathematical knowledge. This shows the urgent need for LLM assistants tailored to cryptography research and development. We demonstrate that, by using CryptoQA, LLMs can be fine-tuned to exhibit better performance on cryptographic tasks.

cs.CR↗

arXiv2025

REFLECTing SPERET: Measuring and Promoting Ethics and Privacy Reflexivity in Eye-Tracking Research

The proliferation of eye tracking in high-stakes domains - such as healthcare, marketing and surveillance - underscores the need for researchers to be ethically aware when employing this technology. Although privacy and ethical guidelines have emerged in recent years, empirical research on how scholars reflect on their own work remains scarce. To address this gap, we present two complementary instruments developed with input from more than 70 researchers: REFLECT, a qualitative questionnaire, and SPERET (Latin for "hope"), a quantitative psychometric scale that measures privacy and ethics reflexivity in eye tracking. Our findings reveal a research community that is concerned about user privacy, cognisant of methodological constraints, such as sample bias, and that possesses a nuanced sense of ethical responsibility evolving with project maturity. Together, these tools and our analyses offer a systematic examination and a hopeful outlook on reflexivity in eye-tracking research, promoting more privacy and ethics-conscious practice.

cs.HC↗

arXiv2025

HAGI++: Head-Assisted Gaze Imputation and Generation

Mobile eye tracking plays a vital role in capturing human visual attention across both real-world and extended reality (XR) environments, making it an essential tool for applications ranging from behavioural research to human-computer interaction. However, missing values due to blinks, pupil detection errors, or illumination changes pose significant challenges for further gaze data analysis. To address this challenge, we introduce HAGI++ - a multi-modal diffusion-based approach for gaze data imputation that, for the first time, uses the integrated head orientation sensors to exploit the inherent correlation between head and eye movements. HAGI++ employs a transformer-based diffusion model to learn cross-modal dependencies between eye and head representations and can be readily extended to incorporate additional body movements. Extensive evaluations on the large-scale Nymeria, Ego-Exo4D, and HOT3D datasets demonstrate that HAGI++ consistently outperforms conventional interpolation methods and deep learning-based time-series imputation baselines in gaze imputation. Furthermore, statistical analyses confirm that HAGI++ produces gaze velocity distributions that closely match actual human gaze behaviour, ensuring more realistic gaze imputations. Moreover, by incorporating wrist motion captured from commercial wearable devices, HAGI++ surpasses prior methods that rely on full-body motion capture in the extreme case of 100% missing gaze data (pure gaze generation). Our method paves the way for more complete and accurate eye gaze recordings in real-world settings and has significant potential for enhancing gaze-based analysis and interaction across various application domains.

cs.HC↗

arXiv2025

Interactive Expressive Motion Generation Using Dynamic Movement Primitives

Our goal is to enable social robots to interact autonomously with humans in a realistic, engaging, and expressive manner. The 12 Principles of Animation are a well-established framework animators use to create movements that make characters appear convincing, dynamic, and emotionally expressive. This paper proposes a novel approach that leverages Dynamic Movement Primitives (DMPs) to implement key animation principles, providing a learnable, explainable, modulable, online adaptable and composable model for automatic expressive motion generation. DMPs, originally developed for general imitation learning in robotics and grounded in a spring-damper system design, offer mathematical properties that make them particularly suitable for this task. Specifically, they enable modulation of the intensities of individual principles and facilitate the decomposition of complex, expressive motion sequences into learnable and parametrizable primitives. We present the mathematical formulation of the parameterized animation principles and demonstrate the effectiveness of our framework through experiments and application on three robotic platforms with different kinematic configurations, in simulation, on actual robots and in a user study. Our results show that the approach allows for creating diverse and nuanced expressions using a single base model.

cs.RO↗

arXiv2025

ToM-SSI: Evaluating Theory of Mind in Situated Social Interactions

Most existing Theory of Mind (ToM) benchmarks for foundation models rely on variations of the Sally-Anne test, offering only a very limited perspective on ToM and neglecting the complexity of human social interactions. To address this gap, we propose ToM-SSI: a new benchmark specifically designed to test ToM capabilities in environments rich with social interactions and spatial dynamics. While current ToM benchmarks are limited to text-only or dyadic interactions, ToM-SSI is multimodal and includes group interactions of up to four agents that communicate and move in situated environments. This unique design allows us to study, for the first time, mixed cooperative-obstructive settings and reasoning about multiple agents' mental state in parallel, thus capturing a wider range of social cognition than existing benchmarks. Our evaluations reveal that the current models' performance is still severely limited, especially in these new tasks, highlighting critical gaps for future research.

cs.CL↗

arXiv2025

The Overcooked Generalisation Challenge: Evaluating Cooperation with Novel Partners in Unknown Environments Using Unsupervised Environment Design

We introduce the Overcooked Generalisation Challenge (OGC) - a new benchmark for evaluating reinforcement learning (RL) agents on their ability to cooperate with unknown partners in unfamiliar environments. Existing work typically evaluated cooperative RL only in their training environment or with their training partners, thus seriously limiting our ability to understand agents' generalisation capacity - an essential requirement for future collaboration with humans. The OGC extends Overcooked-AI to support dual curriculum design (DCD). It is fully GPU-accelerated, open-source, and integrated into the minimax DCD benchmark suite. Compared to prior DCD benchmarks, where designers manipulate only minimal elements of the environment, OGC introduces a significantly richer design space: full kitchen layouts with multiple objects that require the designer to account for interaction dynamics between agents. We evaluate state-of-the-art DCD algorithms alongside scalable neural architectures and find that current methods fail to produce agents that generalise effectively to novel layouts and unfamiliar partners. Our results indicate that both agents and curriculum designers struggle with the joint challenge of partner and environment generalisation. These findings establish OGC as a demanding testbed for cooperative generalisation and highlight key directions for future research. We open-source our code.

cs.LG↗

arXiv2025

Gaze3P: Gaze-Based Prediction of User-Perceived Privacy

Privacy is a highly subjective concept and perceived variably by different individuals. Previous research on quantifying user-perceived privacy has primarily relied on questionnaires. Furthermore, applying user-perceived privacy to optimise the parameters of privacy-preserving techniques (PPT) remains insufficiently explored. To address these limitations, we introduce Gaze3P -- the first dataset specifically designed to facilitate systematic investigations into user-perceived privacy. Our dataset comprises gaze data from 100 participants and 1,000 stimuli, encompassing a range of private and safe attributes. With Gaze3P, we train a machine learning model to implicitly and dynamically predict perceived privacy from human eye gaze. Through comprehensive experiments, we show that the resulting models achieve high accuracy. Finally, we illustrate how predicted privacy can be used to optimise the parameters of differentially private mechanisms, thereby enhancing their alignment with user expectations.

cs.HC↗