Computer Vision
Understanding Autonomous Driving Datasets by Describing Differences between Image Subsets in Natural Language
Julian Truetsch, Felix Hauser, Christoph Stiller et al. (+1 more)
Understanding the composition of large-scale autonomous driving
datasets is essential for safety, robustness, and reliable operation across domains. For example, domain shift between locations could lead to the operating environment being misaligned with the training data, resulting in potentially dangerous performance degradation. Yet, existing data analysis pipelines largely rely on metadata, predefined labels, or manual inspection, which provide limited semantic insight or do not scale. This paper studies set difference captioning: given two subsets of images, the goal is to produce a natural-
language hypothesis describing differences between the target and reference set. Building on a two-stage formulation, we adapt the method to autonomous driving by focusing on object-centric patches derived from object detection, which simplifies aggregation and enables attribution of differences to specific object instances or categories. To evaluate this setting in-domain, we introduce a new
benchmark, AD-Diff Bench. Low-concentration experiments assess the suitability of set-difference-captioning approaches to sparse, real-world differences. We restrict our experiments to open-weight models to support reproducibility and ease of deployment. The proposed
benchmark and analysis provide a step towards practical, human-interpretable
dataset introspection for autonomous driving
datasets. Our implementation and
benchmark dataset are available at https://github.com/KIT-MRT/AD-Diff
Read the full abstract ▾
All authors Julian Truetsch, Felix Hauser, Christoph Stiller, Frank Bieder
Computer Vision
CoFiE: Coarse-to-Fine Evidence Selection for Efficient Streaming Video Understanding
Jing Jiang, Yiran Ling, Ruonan Li et al. (+2 more)
Streaming
video understanding requires
Vision Language Models (VLLMs) to process growing
video streams and answer user questions under tight latency constraints. Existing methods improve efficiency through
token pruning and
memory-bank schemes, but mainly reduce visual
tokens after visual encoding. Consequently, downstream
token pruning alone cannot substantially reduce end-to-end latency because the expensive frame encoding cost has already been incurred. We propose CoFiE, a Coarse-to-Fine Evidence Selection framework that decouples evidence selection into a coarse, query-agnostic filtering stage before the
vision encoder and a fine, query-specific refinement stage during LLM prefill. CoFiE introduces Novelty-Guided Frame Filtering to retain visually distinctive candidate frames and Query-Specific Evidence Refinement to select the frames most relevant to the user query. This design removes substantial redundancy before frame encoding while preserving query-specific refinement once semantic information becomes available. Experiments show that CoFiE establishes a new state-of-the-art accuracy-efficiency trade-off across multiple
video understanding
benchmarks, reaching 78.86% accuracy on StreamingBench and 68.72% on OvO-Bench, with improvements of up to 3.15% over prior methods. Even with up to 80% evidence-frame filtering, CoFiE outperforms strong open-source
multimodal models while improving end-to-end inference latency by up to 2.54 times.
Read the full abstract ▾
All authors Jing Jiang, Yiran Ling, Ruonan Li, Dimitrios Stamoulis, Jie Liu
Information Retrieval
LLM4AIGQ: LLM-based AI Guidance Query Generation Framework for Multi Interest Mining
Xiangchen Pan, Jiayi Xu, Jing Wang et al. (+2 more)
Guidance queries stimulate user consumption by extracting preferences to provide search queries with guidance value, playing a crucial role in the e-commerce field. Traditional AI-generated queries (AIGQ) generation primarily relies on a two-stage "Query-to-AI-Generated-Query" (Q2AIGQ) association paradigm, first recalling user primary search queries from user profiles, historical behavior sequences, item-side information, and the current query through multi-path
retrieval, then generalizing AIGQ via rule-based methods. This approach suffers from semantic drift due to information cascade loss; additionally, primary search query derivation heavily depends on "user-item" co-occurrence relationships, lacking exploration of user multi-interests, resulting in guidance queries with low value and mismatched purchase intent. To address the expressive limitations of traditional co-occurrence-based
retrieval, we propose LLM4AIGQ, an LLM-based solution for generating AI guidance queries tailored to users' multi-interests. This approach segments user interests by integrating user profiles and historical interaction sequences, infers specific consumption intents for each sub-interest, and subsequently generates corresponding AIGQ. In terms of model training, we employ a post-training pipeline comprising Supervised Fine-Tuning (SFT), Reinforcement Learning (RL), and Direct Preference Optimization (DPO) to enhance the model's capability in generating AIGQ. We also introduce a multi-level reward design to satisfy the requirements of multi-objective optimization and long-chain
reasoning in practical applications. Regarding deployment, we adopt a nearline-generation and online-read architecture to meet latency constraints. Extensive experimental analyses demonstrate that our model achieves robust performance in both offline evaluations and online A/B tests.
Read the full abstract ▾
All authors Xiangchen Pan, Jiayi Xu, Jing Wang, Xing Fang, Lingyun Zhu
Computer Vision
Do Video Generators Track the World Across Segments? A Benchmark and Method for World-State Reasoning in Video Continuation
Yingmao Miao, Pengfei Zhang, Chaoran Xu et al. (+5 more)
Video generators build long
videos by composing shorter parts, either by generating segments one after another or by autoregressively extending chunks. Each new part usually depends on
memories of historical observations, such as recent frames, selected key frames,
memory banks, or cached features. These
memories preserve visible evidence from the past, but current generators do not reliably turn such evidence into a world-state interface: what holds in the
video world after previous actions and how it should change under the next
prompt. A past frame remains valid history, but it may not describe the state needed by the next segment; some states must instead be inferred from occluded or implicit changes rather than copied from a directly observed frame. This creates a simple but overlooked question for
video continuation: given a previous
video, its
prompt, and a new
prompt, can a model generate a continuation that reflects the state determined by both the historical
video and the new
prompt? To answer this question, we introduce Statebench, a
benchmark that targets this gap by testing continuations over three state categories: past-visible states, occluded-process states, and complex-transition states. We further propose Stateagent, which explicitly maintains an entity-state representation, updates it under the new
prompt, grounds the predicted post-action state as a future end frame, and renders the next
video. Experiments show that our method improves controlled
video continuation by raising the all-case state score (SCS-All) from 45.2 to 69.3, and also benefits story generation at the one-minute scale. Code is avaliable at https://github.com/AMAP-ML/StateAgent.
Read the full abstract ▾
All authors Yingmao Miao, Pengfei Zhang, Chaoran Xu, Meng Yu, Jing Tang, Xiangxiang Chu, Chao Shen, Chenhao Lin
Databases
Convolution Sum-Product Queries
Kyle Deeds, Timo Camillo Merkl and Dan Suciu
We study query evaluation for an extension of sum-product queries (SPQ) that allows atoms with linear combinations of variables (e.g. $A(2X+3Y-Z, Y+Z)$), which we call convolution sum-product queries (CSPQs). These queries arise in both practical settings (e.g. image processing workloads) and theoretical ones (e.g. the $(\min,+)$-convolution and $k$-SUM
conjectures), and capture SPQs with added linear equality constraints. While prior work has considered the impact of linear in- and dis-equality constraints on query evaluation, the techniques developed in that setting are asymptotically sub-optimal for CSPQs. To address this, we describe several evaluation
algorithms for CSPQs that leverage linear
algebra techniques like rank analysis, quotient spaces, and variable substitution. First, we adapt Worst-Case Optimal Joins to CSPQs, achieving a runtime similar in spirit to that for conjunctive queries. Then, we extend the definition of tree decompositions (TDs) to CSPQs, and describe a factorized execution. In this extension, linear combinations are first-class citizens and play the same role as variables in traditional TDs. We define three width measures that bound the complexity of this execution with respect to the size of the domain values, of the active domain, and of the relation's support. Lastly, we show how these methods can be applied to arbitrary fields beyond the rationals.
Read the full abstract ▾
Logic in Computer Science
The Stochastic Target Discounted-Sum Problem
Nathalie Bertrand, Pranav Ghorpade, Senthil Rajasekaran et al. (+2 more)
The target discounted-sum problem (TDS) asks, given a finite integer alphabet $\Sigma$, a rational discount factor $\lambda$, and a rational target $t$, whether some infinite sequence over $\Sigma$ has discounted sum exactly $t$. This problem remains open and underlies several open questions in automata theory, games, and Markov decision processes. We introduce and solve its
stochastic counterpart, the
stochastic target discounted-sum problem, which replaces existence by computation of the probability. We show that the probability that a random sequence generated by a finite Markov chain has discounted sum $t$ is rational and computable in pseudo-
polynomial time. We further show how to decide, in
polynomial time, whether the discounted-sum distribution of a Markov chain is atomless, and how to approximate to an arbitrary precision the probability that the discounted sum exceeds a rational threshold. Our techniques for the
stochastic TDS problem allow us to make progress on TDS objectives in
stochastic games, which are known to be as hard as the TDS problem. Restricting the maximizing player to finite-
memory strategies, while allowing the minimizing player to use arbitrary strategies, we reduce the value problem and the synthesis problem to corresponding problems for safety objectives in
stochastic games. This yields computable optimal values and deterministic optimal strategies with pseudo-polynomially bounded
memory for
stochastic games, and results in pseudo-
polynomial-time
algorithms for special cases of Markov decision processes and deterministic two-player games.
Read the full abstract ▾
All authors Nathalie Bertrand, Pranav Ghorpade, Senthil Rajasekaran, Sasha Rubin, Moshe Y. Vardi
Computer Vision
ARCOS: Zero-shot Boundary Localization for Corneal Layer Segmentation Across Optical Coherence Tomography Devices
Nuno Vivas Brás, Benjamin Memmi, Maëlle Bouhassane et al. (+4 more)
Accurate
segmentation of corneal layers in optical coherence
tomography (OCT) is essential for quantitative assessment of corneal morphology, including layer thickness and structural changes associated with disease or surgery. However, automatic
segmentation remains challenging because corneal interfaces are thin, affected by speckle noise, and variable across acquisition devices. In this work, we propose ARCOS, a patch-based zero-shot boundary localization framework for corneal layer
segmentation in clinical anterior-segment OCT images. Rather than performing conventional region classification, the method predicts boundary heatmaps for the main corneal interfaces from overlapping native-resolution patches. Patch-level predictions are stitched across the full B-scan and converted into boundary locations to obtain continuous, anatomically ordered layer
segmentations. The network combines multi-scale feature fusion with a self-conditioned refinement module that uses intermediate boundary information to improve local heatmap predictions while preserving spatial detail. The method was evaluated on clinical OCT images acquired from multiple devices and compared with representative
segmentation baselines using boundary localization and derived thickness metrics. The proposed method achieved an off-by-one boundary localization accuracy of 95.1% and a mean absolute boundary error of 0.514 pixels on the matched-device test set. In zero-shot cross-device evaluation, it maintained an average off-by-one accuracy of 84.3% and a mean absolute boundary error of 0.855 pixels across unseen acquisition devices, outperforming the baseline models. Thickness estimates derived from the predicted boundaries showed low error across corneal regions, supporting the method's use for quantitative corneal OCT analysis.
Read the full abstract ▾
All authors Nuno Vivas Brás, Benjamin Memmi, Maëlle Bouhassane, Cristina Georgeon, Vincent Borderie, Karsten Plamann, Anatole Chessel
Machine Learning
Out-of-Distribution Generalisation with Sequence Models in Offline Multi-Agent Reinforcement Learning
Oussama Hidaoui, Omer Ebead, Ulrich Armel Mbou Sob et al. (+14 more)
Generalising to unseen tasks remains a fundamental challenge in offline multi-
agent reinforcement learning (MARL). In this work, we present a principled analysis of zero-shot task generalisation in the offline setting and conduct an extensive empirical investigation into the scaling behaviour governing task diversity,
dataset size, and network capacity. To facilitate this study, we extend offline sequence modelling architectures to handle multi-task observation and action spaces alongside variable
agent counts across tasks. Our primary finding is that scaling task diversity—rather than sheer
dataset size is the dominant factor in achieving robust zero-shot transfer. Through large-scale experiments across four challenging environments (Connector, RWARE, SMAX, and LBF), we demonstrate that our multi-task approach achieves a mean improvement of 3.2x on held-out test tasks compared to single-task models and consistently outperforms strong behaviour cloning baselines. These results suggest that the development of generalisable MARL
agents should prioritise the diversity of the training distribution with varying numbers of
agents, providing a roadmap for scaling offline MARL effectively.
Read the full abstract ▾
All authors Oussama Hidaoui, Omer Ebead, Ulrich Armel Mbou Sob, Siddarth Singh, Juan Claude Formanek, Felix Chalumeau, Omayma Mahjoub, Sasha Abramowitz, Ruan John de Kock, Wiem Khlifi, Louay Ben Nessir, Simon Verster Du Toit, Daniel Rajaonarivonivelomanantsoa, Asim Awad Osman, Arnol Manuel Fokam, Refiloe Shabe, Arnu Pretorius
Computers and Society
WebXR and Commercial Game Engines for the Metaverse: A Socio-Technical Analysis of Openness, Interoperability, and Sustainability
Luca Turchet and Michel Buffa
The Metaverse is often framed as a persistent, interoperable, and
embodied network of virtual and augmented environments. Yet, most contemporary XR applications are developed through commercial game engines and distributed through proprietary app stores, creating tensions between openness and platform dependency. This paper critically examines open WebXR technologies with conventional commercial game-engine pipelines, with particular
attention to XR hardware, software architectures, developer workflows, governance, ethics, interoperability, and sustainability. We argue that WebXR may provide a viable route toward a more accessible, device-independent, and institutionally sustainable Metaverse, especially for education, research, cultural heritage, prototyping, and public-interest applications. At the same time, commercial engines remain advantageous for graphically intensive, low-latency, deeply integrated, and large-scale XR products. The paper concludes that the choice should not be framed as WebXR versus engines, but as a continuum: WebXR is preferable when accessibility, interoperability, low-friction deployment, and long-term maintainability are primary goals, whereas native engines remain preferable when performance, platform-specific hardware access, and production-grade tooling dominate.
Read the full abstract ▾
Human-Computer Interaction
PlanePivoting: Exploration and Optimization of Gaze-Mouse Cursor Alignment for Spatial Object Translation
Jinwook Kim, Sangmin Park, Jihyeon Lee et al. (+2 more)
As XR matures into a ubiquitous computing platform, the disconnect between 2D and 3D input modalities remains a critical barrier to seamless workflow. Frequent transitions between the mouse for 2D precision and hand gestures for 3D manipulation induce significant physical fatigue and cognitive load. To address this, we introduce PlanePivoting, a
multimodal interaction technique that extends standard mouse input into 3D space by leveraging gaze-mouse
alignment. This technique dynamically modulates the translation plane based on the spatial overlap between the gaze and mouse cursor, eliminating the need for physical input modality switching. To systematically explore the foundational design space of gaze-mouse coordination and optimize key variables, we conducted a user study comparing PlanePivoting with a standard 3D Gizmo interface across two translation mapping profiles and two gaze cursor apertures. Results demonstrate that PlanePivoting outperforms the Gizmo on efficiency metrics while maintaining comparable precision and yielding higher subjective satisfaction. This study demonstrates the potential of gaze-mouse
alignment for efficient spatial manipulation between 2D and 3D environments.
Read the full abstract ▾
All authors Jinwook Kim, Sangmin Park, Jihyeon Lee, Sang Ho Yoon, Jeongmi Lee
Computer Vision
Cross-Dataset Transfer and Reliability of Explainable Artificial Intelligence for RhythmFormer Remote Photoplethysmography
Louis Chen and Torbjörn E. M. Nordling
Background. Remote photoplethysmography estimates the cardiovascular pulse from facial
video, and its explanations have rested on inspecting heatmaps rather than on quantitative evidence about where a model reads it. We quantified the explanations and asked whether such explanations transfer between
datasets and track model performance. Method. We trained eight condition-specific RhythmFormer models on NCKU-rPPG, recorded under three illumination levels, speaking, rotation, and cycling, estimated one heart rate per 5.12-second clip, and set them beside a UBFC-rPPG reproduction. Raw
attention, rollout,
attention flow, and Beyond Intuition were assessed by skin coverage and the Salience-guided Faithfulness Coefficient (SaCo). Results. Beyond Intuition ranked highest on both
datasets, at median coverage 0.789 and SaCo 0.837 on Static level 3 against 0.826 and 0.917 on UBFC-rPPG; lower ranks differed. Within one participant of one condition, neither measure was related to a clip's heart-rate error, waveform correlation, or signal-to-noise ratio on either
dataset: 186 of the 252 coefficients fell below $|\rho|=0.10$ and 28 reached $p<0.05$ against the 13 expected by chance. Across the eight scenarios only Beyond Intuition's coverage followed the three performance measures, at $\rho=-0.43$, $+0.57$, and $+0.43$, while the
attention-only methods' SaCo ran opposite to each. It failed at 40 lux alone, its median coverage falling to 0.180 and its median SaCo to $-0.178$, whereas motion degraded the estimates far more without such a drop. Conclusions. Skin coverage and SaCo carry information complementary to the performance measures rather than a proxy for them: attributing to the skin does not guarantee an accurate estimate. What an attribution reveals about a condition is where the model looks rather than how faithfully its map is ordered.
Read the full abstract ▾
Machine Learning
Extracting Forgotten Prompts from Targeted Unlearned Models
Au Ashley Hoi-Ting, Meghdad Kurmanji, William F. Shen et al. (+2 more)
Recent unlearning methods (e.g. NPO, DPO, LUNAR) make use of refusal
alignment to suppress forgotten data. However, it has been shown that refusal responses might leave traces of unlearning, and recent attacks have been able to successfully recover some of the unlearned knowledge. In this paper, we uncover a new vulnerability. Existing attacks typically assume that the forgotten
prompts are already known to the adversary and focus on recovering their answers. However, we show that the forgotten
prompts themselves can be extracted by using the retained data and black-box access to the model. Our attack, Targeted Active Search (TAS), first identifies the forgotten entities by constructing canonical templates and entity pool, and selectively querying the model using the most informative template-entity pair under a limited query budget. Once the entities are identified, TAS instantiates
prompt templates with those entities to probe the unlearned model and reconstruct the forgotten
prompts. Experiments across three unlearning methods with three
datasets and three LLMs shows that TAS recovers the forgotten entity with $100\%$ accuracy and reconstructs up to $95\%$ of forgotten
prompts, all while using up to $99.7\%$ fewer queries than naive probing.
Read the full abstract ▾
All authors Au Ashley Hoi-Ting, Meghdad Kurmanji, William F. Shen, Nicholas D. Lane, Ligang He
Human-Computer Interaction
Point&Spawn: Mid-Air Reference-Free Object Instantiation Using Gaze and Hand Gestures in Extended Reality
Jihyeon Lee, Ken Pfeuffer, Jinwook Kim et al. (+1 more)
Mid-air object instantiation in XR requires users to specify a 3D position without spatial references, such as surfaces or existing objects. We present Point&Spawn, a staged pipeline for pre-instantiation position specification through Direction Setting, Depth Setting, and Position Refinement within a continuous gesture flow. We evaluated six controller-free techniques combining Gaze or Non-Dominant Hand (NDH) direction setting with Ray Intersection, Relative Gain, or Drag&Hold depth setting in a user study (N=24) across Near and Far spawn depths. Relative Gain and Drag&Hold yielded faster and more accurate spawning, lower workload, higher usability, and greater preference than Ray Intersection. The shoulder-referenced NDH ray improved speed and coarse accuracy, whereas the viewpoint-based Gaze ray reduced hand movement with comparable final accuracy. Farther spawn depth imposed greater temporal costs as well as Gaze and accuracy costs with Ray Intersection. These findings offer empirical guidance for designing direction and depth control in spawning in XR.
Read the full abstract ▾
All authors Jihyeon Lee, Ken Pfeuffer, Jinwook Kim, Jeongmi Lee
Machine Learning
Local Updates, Global Learning (LUGL): Playing Games with non-incremental Learners
David Milec, Spyridon Samothrakis, Michael Fairbank et al. (+1 more)
The dominance of Neural Networks (NNs) in RL is partially due to their incremental learning capability, which naturally suits the online, non-stationary nature of self-play training. However, gradient-boosted trees like LightGBM are widely recognised as the state of the art for tabular data in supervised learning, often outperforming NNs in accuracy and efficiency. Game states are inherently tabular—discrete actions, categorical card identities, structured board positions—which makes them an ideal candidate for tree-based methods. We introduce LUGL (Local Updates, Global Learning), a framework that decouples data collection from model fitting, enabling non-incremental learners such as GBTs to operate in RL settings where they would otherwise fail due to distributional shift. LUGL alternates between a local updates phase, where the
agent plays self-play games and accumulates tabular updates (Q-values, V-values, policies, or regret values) in a finite table, and a global learning phase, where the table is used to train a function approximator that generalises to unseen states before the table is reset. We test our approach in four standard perfect-information games (Tic-tac-toe, Connect-4, Othello, and Hex) and five imperfect-information games (Kuhn's poker, Leduc Hold'em, Liar's Dice, Goofspiel, and Flop5 Hold'em), and show that our results are competitive with or superior to DQN and DeepCFR. Our experiments demonstrate that the community's strong bias towards NNs in game-playing may be unwarranted, since LightGBM-based
agents achieve competitive or superior performance across all tested
benchmarks.
Read the full abstract ▾
All authors David Milec, Spyridon Samothrakis, Michael Fairbank, Dennis J.N.J. Soemers
Cryptography and Security
Security and Privacy in the Musical Metaverse: Threat Analysis and Design Implications
Luca Turchet and Michał Kłosinski
The Musical Metaverse (MM) introduces immersive, real-time environments for collaborative musical interaction, characterized by ultra-low-latency constraints, continuous
multimodal data streams, and heterogeneous devices. These properties create a distinctive security and
privacy landscape that differs significantly from conventional XR or multimedia systems. This paper presents a multi-layer threat analysis of MM ecosystems, identifying key assets including live musical content, expressive interaction data, identity and session metadata, and intellectual property. Threats are analyzed across network, application, data/AI, device, intellectual property rights, and social layers, with particular
attention to risks arising from expressive and neurophysiological data, which enable inference, re-identification, and potential
privacy violations. We describe a stakeholder-driven survey involving 14 participants from 13 organizations, revealing that neurophysiological data leakage and real-time stream disruption are perceived as the most critical risks, followed by intellectual property infringement and avatar impersonation. We further evaluate the suitability of existing security protocols under strict latency constraints, showing that conventional approaches such as TLS over TCP are often incompatible with real-time musical interaction, while lightweight, stream-oriented mechanisms (e.g., SRTP, DTLS) provide a more suitable balance between security and performance. Based on these findings, we derive a set of design guidelines for MM systems, emphasizing latency-aware security, differentiation of interaction paths, data minimization, and edge-centric processing. The results support a security-by-design approach that enables trust and compliance without compromising real-time performance.
Read the full abstract ▾
Computer Vision
Rethinking 3D Noise: Learning 3D-Aware Video Priors via Optimization-Free Morphological Perturbations
Onat Şahin, Mohammad Altillawi, George Eskandar et al. (+2 more)
3D scene representations like NeRF and 3D Gaussian Splatting (3DGS) suffer severe artifacts in sparse-view settings. Recent generative 3D artifact fixers attempt to address this, but rely on paired corrupted and clean renders requiring costly, per-scene reconstructions across varying view configurations. While 2D image augmentations act as instant regularizers, no explicit equivalents exist for 3D representations to preserve spatial consistency across views, an essential property for 3D-aware training. We propose 3D Morphological Perturbations as an optimization-free regularizer that preserves spatial consistency. Leveraging explicit 3DGS, we treat each Gaussian as a fundamental building block - analogous to a 2D pixel - and apply perturbations across its morphological parameter space via scale, rotation, and
pruning. Our method eliminates per-scene 3DGS optimization loops from
dataset curation while enabling models to learn stronger
geometric priors than sparse-view baselines in diagnostic ablations conducted on a lightweight
video diffusion sandbox. Scaled to a 14B-parameter
video model via ControlNet, our approach maintains visual fidelity while reducing mean depth error by 12.5% over state-of-the-art image-to-image 3D artifact refiners, ultimately boosting downstream robotics policy success rates by up to 8.0% across 3 of 4 manipulation tasks.
Read the full abstract ▾
All authors Onat Şahin, Mohammad Altillawi, George Eskandar, Carlos Carbone, Ziyuan Liu
Computer Vision
PL-SCEA: Reconfiguring Pretrained Attention for Few-Shot Industrial Anomaly Detection
Xiaoyu Yang, Qixing Wu, Huixian Zhao et al. (+1 more)
Vision Foundation Models (VFMs) provide transferable patch representations for few-shot industrial anomaly detection, but their
attention computation is typically inherited from pretraining objectives centered on semantic aggregation. This creates a potential mismatch:
token relations that support semantic recognition may not adequately expose the localized texture and structural deviations required for anomaly localization. We therefore investigate the hypothesis that the
attention computation of a frozen VFM can be reconfigured as a task-relevant component of anomaly detection. We instantiate this idea with Power-Law Self-Correlation Enhanced
Attention (PL-SCEA), which retains the semantic context of pretrained query-key
attention while constructing
token-adaptive self-correlations over contextualized value features. Positive-correlation filtering and power-law reweighting then emphasize relations that are salient relative to each
token's relational background, without introducing additional trainable
attention projections. The resulting features are modeled by a lightweight variational autoencoder that provides a fixed-size reconstruction-based representation of category-specific normality. The two stages serve complementary roles:
attention reconfiguration shapes how local relational deviations are represented, while reconstruction-based modeling converts deviations from learned normality into anomaly scores. Across MVTec AD and VisA, the complete framework achieves competitive image-level detection and consistently strong pixel-level localization across the evaluated few-shot settings. Ablations further show that PL-SCEA improves localization with either the VAE or a
memory bank under the tested setting. These results support the view that task-aligned
attention reconfiguration can improve the anomaly-localization capability of frozen pretrained representations.
Read the full abstract ▾
All authors Xiaoyu Yang, Qixing Wu, Huixian Zhao, Changlong Jin
Computation and Language
Enhancing Financial Question Answering: A Novel Benchmark Dataset of Banks' financial statements
Arianna Miola, Bruno Spaccavento, Lorenzo Silotto et al. (+2 more)
The comparative analysis of banks' financial statements poses significant challenges for automated question answering systems due to their complexity, substantial length, technical
language, and inhomogeneity of both textual and numerical content across different jurisdictions and institutions. We introduce FinRAG-QA, a novel
benchmark dataset for financial question answering, which comprises 999 practitioner-curated questions on 10 standardised indicators, grounded in 209 annual and Pillar 3 reports from 24 major European and U.S. banks spanning 2019-2023. Unlike prior financial QA
benchmarks, which centre on U.S. filings and single-institution analysis, FinRAG-QA targets cross-institutional
retrieval over documents averaging 198k words, longer than any existing financial QA resource. On this
benchmark we evaluate a multi-stage RAG pipeline and isolate the contribution of each component. Contextual chunk enrichment combined with a
retrieval-optimised
embedding model raises NDCG@10 from 0.322 to 0.710; conditional on the ground truth being retrieved, a
reasoning-optimised generator raises answer accuracy from 44.6% to 79.0% (+34.4 percentage points), at roughly 20x the generation latency. We further show that cross-encoder reranking degrades
retrieval when the first-stage ranking is already strong, and that a single top-ranked chunk outperforms larger contexts at generation time. Experiments were run in late 2024-early 2025 with the models available at that time.
Read the full abstract ▾
All authors Arianna Miola, Bruno Spaccavento, Lorenzo Silotto, Marco Bianchetti, Luca Cagliero
Computation and Language
The Impact of Synthetic Data Augmentation on Discourse-Pragmatic Function Classification
Sara Sorahi, Kevin Tang and Reza Kazemian
Synthetic data augmentation has become a common strategy for addressing class imbalance in NLP, but most approaches focus on the quantity and diversity of generated examples rather than their
geometric relationship to real training data. We investigate this question in the context of discourse pragmatic function classification, a task where data sparsity is a structural feature rather than a collection artefact. Using 410 manually annotated instances of the English word look drawn from the British National Corpus, spanning four functions:
Attention Signal, Directive, Discourse Marker, and Interjection. We generate synthetic training examples with Llama 3.1 and partition them by their cosine distance from real training data in RoBERTa
embedding space. We compare six training conditions that differ in the placement of synthetic examples relative to the empirical decision boundary, while holding augmentation quantity constant across conditions. All augmented conditions improve macro F and accuracy over the real only baseline, but core proximal examples (NEAR) yield the largest gains in macro F (0.113), while a distance balanced mix achieves the highest accuracy (0.748). No condition improves AUC, indicating that augmentation shifts the decision boundary rather than improving the model's underlying probability estimates. These findings suggest that where synthetic examples land in representation space matters as much as how many are generated, with implications for low resource pragmatic classification more broadly.
Read the full abstract ▾
Computational Engineering
Calibration of neural viscoelastic models via full-field data
Brain M. Riemer, Markus Kästner and Karl A. Kalina
We propose an unsupervised learning framework for calibrating a physics-augmented neural network (PANN) for small-strain viscoelasticity via full-field data. It only requires quantities that are directly accessible in real experiments for training, namely global reaction forces and surface displacements. The underlying PANN is embedded in the generalized standard materials theory, in which two scalar-valued potentials render the constitutive model thermodynamically consistent by construction, while invariant-based representations of the free energy and the dual dissipation potential additionally ensure material
symmetry. Considering a thin specimen under the plane stress assumption, we formulate a constrained optimization problem based on the
equilibrium gap method in combination with quasi-Newton optimizers and automatic differentiation. Thereby, the unknown out-of-plane strain follows from the plane stress condition and the
evolution of the internal variables is captured by an implicit time integration scheme. The resulting system of nonlinear equations is solved via a local Newton iteration at quadrature point and time step. To drastically reduce the computational cost of training, the backward adjoint method is employed to compute the gradient of the target loss, instead of backpropagating through all Newton iteration steps. The proposed framework is demonstrated for synthetic data, including noisy displacements and forces, showing excellent agreement across a wide range of deformation rates and load paths.
Read the full abstract ▾
Formal Languages and Automata
Relative Prime Factorization and Finite-State Presentations under Fixed Finite-Monoid Observation
Takayuki Kuriyama
Let $L\subseteq\Sigma^*$ and fix a morphism $h:\Sigma^*\to M$ into a finite monoid. We study exact factorization and canonical presentation in the relative syntactic congruence $\theta_{L,h}:=\equiv_L\cap\ker h$. We separate unique factorization from finite direct presentation. An exhaustively computer-checked $36$-element quotient has a unique exact
prime factorization for every live non-unit class, yet its valid
prime-return rules contain an infinite family, so unique factorization does not imply the finite relative presentation property (FRP), even for a finite quotient. We lift the same defect to a nonregular context-free
language with an infinite relative quotient and finite
prime spectrum. To isolate the obstruction, we introduce the finite-state relative presentation property (FSRP), in which canonical valid right-hand-side
languages are represented by finite residual controllers, and prove $\mathrm{FRP}\subsetneq\mathrm{FSRP}$. We then introduce
prime-target left-division determinism (PTLD), which implies unique exact factorization, tail exactness, tail determinism, and a quadratic bound on valid rules. A nonregular deterministic context-free example with a finite group observer satisfies PTLD while lying outside every fixed $(k,\ell)$-substitutable class. Finally, for fixed $h$ we give a strong positive-data learner for the canonical PTLD presentation with
polynomial-time hypothesis updates and a finite characteristic sample, together with a limit reconstruction of the canonical FSRP controller from weakly behaviorally correct CFG-valued learners.
Read the full abstract ▾
Computer Vision
Tree-VQ: Progressive Image Compression from Pretrained Vector Quantizers
Mingming Ma, Xinkun Wang, Tianyi Xu et al. (+3 more)
Progressive image compression requires a single embedded representation whose received prefixes can be decoded without re-encoding the source. Modern vector-quantized (VQ) image models provide strong discrete endpoint representations, but conventional flat codeword indices do not define meaningful intermediate states for a neural decoder. We present Tree-VQ, a post-hoc conversion of a pretrained flat VQ tokenizer into a fine-grained, arbitrary-prefix progressive representation while preserving its encoder assignments and every learned leaf vector. The key idea is to organize the original codebook into a balanced binary hierarchy, associate explicit representations with internal nodes, and transmit branch decisions in depth-major order. Consequently, once the image header is available, every payload prefix uniquely specifies a valid latent state: each additional branch bit refines exactly one
token, and transmission can therefore be truncated at essentially any payload position rather than only at a small number of stage boundaries. We further adapt one shared decoder on the complete-depth and mixed-depth latent states encountered under such arbitrary truncation, making these densely spaced prefixes useful for reconstruction rather than merely syntactically decodable. On Kodak, Tree-VQ achieves a DISTS-based BD-rate saving of 52.1% relative to ProGIC, while exposing thousands of valid arbitrary-prefix operating points from a single embedded bitstream.
Read the full abstract ▾
All authors Mingming Ma, Xinkun Wang, Tianyi Xu, Qingyu Luo, Fu Li, Yi Niu
Computer Vision
Stabilizing Camera-Controlled Novel View Synthesis at Inference Time
Prajwal Singh, Arjun Badola, Seema Kumari et al. (+2 more)
Training-free, camera-controlled novel view synthesis from a single image using pre-trained
video diffusion models often becomes unstable under large camera motion and long generation horizons. Existing approaches commonly combine several inference-time components, making it unclear which design choices are most important for stability. This work identifies the amount of camera motion handled by each
diffusion pass as the main factor. CamTrol++ splits the camera path into passes that each cover at most a fixed rotation and generates them one after another, so that every pass stays within the range of motion the backbone handles well. A matched study, in which every setting is scored on the same poses of the same $60^\circ$ orbit, shows that
geometric error grows steadily with the arc per pass, while a single SVD pass both loses geometry and tends to freeze the camera. Under one shared camera path, CamTrol++ achieves the lowest epipolar error and the best held-out 3D Gaussian Splatting reconstruction among CamTrol, NVS-Solver and WAVE, and reaches the quality of CamTrol with the same pass structure at least $3.2\times$ faster. On real RealEstate10K motion it matches CamTrol at half the run time (NVS-Solver is more faithful but $8\times$ slower). The same procedure also improves 3D reconstruction with a second
video backbone, LTX-
Video. Geometry-constrained spatial
attention, color anchoring at pass boundaries and a registration-free warping pipeline that makes warping $6.1\times$ faster complete the framework. The method remains effective for 56-frame generation and under controlled depth corruption. These results show that careful control of camera motion at inference time can substantially improve the stability of camera-controlled novel view synthesis.
Read the full abstract ▾
All authors Prajwal Singh, Arjun Badola, Seema Kumari, Hajime Nagahara, Shanmuganathan Raman
Artificial Intelligence
Analysis of Prompt Engineering for Drug Toxicity Prediction
Mia MacGregor, Aakash Welgamage Don and Mark Bartlett
Clinical trials in the UK can cost up to {\pounds}1.3 million, with approximately 90% drug failure rate. Toxicity is a major contributing factor in drug failure. Testing is time and cost intensive. In recent years, the use of artificial intelligence has been increasingly explored to aid in the prediction of drug toxicity, with extensive use of large
language models (LLMs). However, LLMs can show considerable variation when minor changes are made to
prompts, which raises concerns about their sensitivity to
prompt engineering.
Prompt engineering is used to optimise a
prompt given to an LLM to generate the desired output. This paper proposes a method to analyse
prompt engineering for drug toxicity prediction. The aim of the paper is to investigate the importance of
prompt phrasing for drug toxicity prediction. LLMs were prompted to identify chemical properties of significance when predicting drug toxicity.
Prompts were constructed to investigate; job role,
prompt structuring, and rule interpretation. LLMs were then used to generate
datasets, using the identified features from initial prompting, which were then passed to machine learning
algorithms. The experiments show that the natural variance which occurs in LLMs outweighs any fine-tuning of
prompts. There were, however, substantial improvements in model performance when using chemoinformatic code to extract features instead of using LLM-generated values. The proposed analysis methodology is applicable to a wide range of
prompt types across different areas of bioinformatics.
Read the full abstract ▾
Computation and Language
</think> Doesn't Stop Reasoning: Analysis of Spurious CoT Termination
Seunghee Koh, Sungjae Choi, Minchan Kwon et al. (+2 more)
Chain-of-thought (CoT)
reasoning improves large
reasoning models (LRMs) on complex tasks but often produces long, redundant traces. Recent training-free early-exit methods shorten these traces by choosing an intermediate point to stop
reasoning. We study one such strategy that injects an end-of-think
token (EoT, </think>) at this point to trigger the
reasoning-to-answering transition, and find that the injected EoT does not always induce a clean answering phase. Answering-phase generation can continue before the model regenerates another EoT, with the span preceding this regenerated EoT scaling with the
reasoning tokens saved by early exit and exhibiting continued
reasoning behavior. We call this spurious CoT termination, where
reasoning-like generation continues into the answering phase. We hypothesize that insufficient
attention to the injected EoT contributes to spurious CoT termination and probe this hypothesis with Exit-
token Attention Biasing (EAB). Across four LRMs, five
benchmarks, and two early-exit methods, increasing
attention to the injected EoT reduces spurious CoT termination and answering-phase length. These results reveal a limitation of controlling LRMs by externally matching their explicit think-block format. Inserting the EoT
token conforms to this format but does not by itself guarantee the intended
reasoning-to-answering transition. Our code is available at https://github.com/Seunghee-Koh/Spurious-CoT-Termination.
Read the full abstract ▾
All authors Seunghee Koh, Sungjae Choi, Minchan Kwon, Sunghyun Baek, Junmo Kim
Social and Information Networks · Cross-listed from physics.soc-ph
Dynamic probabilistic decision networks
V.I. Yukalov and E.P. Yukalova
A new type of decision networks is suggested and its operation is analyzed. The network nodes are represented by intelligent
agents who can denote either some biological beings, like humans, or
neurons of the
brain, or the nodes of artificial intelligence. The specifics of the network are in the following: It is probabilistic in the sense that the choice, accomplished by each
agent, is characterized by the related probability. It is dynamic, with the probabilities varying in time due to the exchange of information between the
agents. It is affective, because the
agents choose between alternatives by taking account of utility as well as of biases and
emotions. In general, it is heterogeneous, being composed of the groups of
agents with different properties, for instance having long-term
memory and short-term
memory. The network dynamics, caused by the information exchange, results in decision error decrease. The network operation is illustrated by the example starting with the Allais paradox, its resolution, and the decision error diminution in the process of decision dynamics with information exchange. Resorting to machine-learning techniques it is possible to regulate the behavior of the network
agents forcing them to choose particular alternatives.
Read the full abstract ▾
Numerical Analysis
A fully globalized solver for discretized inverse elliptic coefficient problems with exact data
Bastian Harrach
We consider finite-dimensional nonlinear inverse problems arising from finite element discretizations of
elliptic inverse coefficient problems such as the Calderón problem with finitely many measurements and unknowns. Such inverse coefficient problems are notorious for their nonlinearity and ill-posedness, and numerical solvers tend to depend strongly on good initial values. In this work, we develop a new locally convergent
algorithm with an explicit residual criterion that ensures convergence to the inverse problem solution, and a globalized variant that is guaranteed to automatically switch to the faster locally convergent
algorithm after finitely many global search steps.
Read the full abstract ▾
Robotics
Local Path Planning and Obstacle Avoidance for an Omnicopter Platform
Mikolaj Helinski, Spilios Theodoulis, Mahmoud Hamandi et al. (+3 more)
Autonomous unmanned aerial vehicles (
UAVs) increasingly operate in cluttered environments where global planners such as RRT* are not directly deployable at control rates. This paper presents a real-time local planning and obstacle avoidance module for an omnidirectional multirotor (omnicopter) by extending the Dynamic Window Approach to six degrees of freedom (6D-DWA). Our method achieves real-time feasibility through (i) local-map voxelisation, (ii) a compact sphere-based approximation of the vehicle geometry, and (iii) adaptive velocity sampling in the 6D search space. To improve reactivity to unknown obstacles, we introduce a context-aware "Agile Mode" that adjusts scoring weights online to trade-off between goal progress, clearance, and heading/facing constraints during evasive manoeuvres. We evaluate our approach in simulation across computational stress tests, dense-waypoint path tracking, and static/unknown obstacle scenarios. Our planner runs consistently within a 0.2s control loop, tracks waypoint-dense global paths with < 0.1m average cross-track error and 13deg average heading error, and avoids collisions in static environments. For unknown obstacle avoidance, Agile Mode achieves 79.3% success for an off-centre obstacle and 41.4% for a centred obstacle, highlighting both the effectiveness of adaptive weighting and remaining limitations in highly constrained geometries.
Read the full abstract ▾
All authors Mikolaj Helinski, Spilios Theodoulis, Mahmoud Hamandi, Abdullah Mohamed Ali, Anthony Tzes, Marija Popovic
Computer Vision
EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders
Xinghao Wang, Dong Li, Wei Yu et al. (+5 more)
Recent advances in text-to-
video (T2V)
diffusion models have demonstrated remarkable generative capabilities, yet their reliance on loosely curated training data raises pressing safety and copyright concerns. Concept erasure offers a principled remedy by removing unwanted semantics from pretrained models while preserving remaining concepts. However, existing approaches typically operate at a coarse granularity misaligned with the fine-grained, distributed nature of concept representations, leading to incomplete removal or degraded generation quality. We argue that
surgical erasure fundamentally requires intervention at the level of monosemantic features, where each unit encodes a single interpretable concept. To this end, we propose EraseSAE, a novel framework that leverages sparse autoencoders to achieve
surgical concept erasure in DiT-based T2V
diffusion models via a principled decompose-attribute-erase pipeline. We first introduce the Partitioned Convolutional Sparse Autoencoder, which decomposes dense spatiotemporal activations into disentangled, interpretable sparse features while preserving spatiotemporal coherence. A contrastive attribution mechanism then contrasts activations from paired
prompts to isolate concept-specific feature kernels. At inference, timestep-resolved spatiotemporal masks derived from the identified kernels confine erasure to regions where the target concept is active, leaving unrelated content intact. Extensive experiments across diverse
diffusion models and concept erasure tasks demonstrate that EraseSAE achieves precise and robust concept removal with minimal quality degradation, substantially outperforming state-of-the-art methods. The code is available at https://github.com/HiDream-ai/EraseSAE.
Read the full abstract ▾
All authors Xinghao Wang, Dong Li, Wei Yu, Yingwei Pan, Tao Gong, Qi Chu, Nenghai Yu, Ting Yao
Computational Complexity
Random Garbage Creates Exponential Quantum Query Separations
Khaled Elbassioni, Rishikesh Gajjala and Saurabh Ray
We prove an exponential separation between standard (XOR)
quantum queries and forward-erasing queries for a Boolean property of injective functions, answering a question of Aaronson~\cite{Aaronson2021Open}. We also obtain an exponential separation for a Boolean promise problem on permutations, where forward in-place queries are
unitary but inverse queries are not supplied. Our construction follows Aaronson's suggestion of appending fixed random garbage to a Simon function. Let $X=\{0,1\}^n$ and $N=2^n$. We use the injection $f_{h,r}(x)=(h(x),x,r_x)$, where $h:X\to X$ is promised to be either a permutation or a Simon function, and $r:X\to X$ is an arbitrary tag table that is reused on every query. The resulting problem is solvable with at most $n+2$ standard queries, but requires $\Theta(\sqrt N)$ forward-erasing queries, matching classical collision sampling.
Embedding these instances into permutations, with the copy of $x$ replaced by a single bit that keeps the
embedding injective, gives a promise problem on permutations of $L=4N^2$ elements that requires $O(\log L)$ standard queries versus $\Theta(L^{1/4})$ forward in-place queries. The lower bounds use the recording method. We replace each query by a nearby contraction that records
cells of a purified tag table. After $T$ queries, tracing out the purified tag table yields a sum of positive-semidefinite contributions, each depending on $h$ at no more than $T$ addresses. A query-by-query hybrid bounds the trace-norm difference between the actual and auxiliary final operators by $O(T/\sqrt N)$. Random permutations and random Simon functions have the same restriction distribution on such a set unless it contains a hidden Simon pair, an event of probability $O(T^2/N)$. Consequently, the actual final states in the two cases remain close unless $T=\Omega(\sqrt N)$, giving the required lower bounds.
Read the full abstract ▾