While large vision-language models excel at optical character recognition, it remains unclear how their underlying vision encoders process text and its meaning, and whether they support anything resembling reading comprehension. We systematically study transformer-based vision encoders, using cross-modal alignment between rendered text and visual concepts together with probing and retrieval-based analyses as a controlled window into these broader capabilities. Our layer-wise analyses span orthographic, grammatical, and semantic organization, and reveal that strong cross-modal mapping emerges only in the deepest layers, after unimodal object and text recognition have already developed, coinciding with a progression from surface-level to semantic representations. Although this hierarchy is strongest in contrastively trained models, larger self-supervised models and text-rich training data partially support similar behaviors without explicit language supervision. Together, our results suggest that increasing model and training-data scale, as well as text-rich training data, can partially compensate for the absence of language supervision.
https://arxiv.org/abs/2610.32463
Image generation models have rapidly advanced, with many models now supporting generation conditioned on multiple reference images and textual instructions. Such models enable a wide range of applications, including entity fusion and attribute transfer across images. However, existing benchmarks for multi-reference image generation rely on limited evaluation dimensions and fail to capture the diverse challenges of the task. To address this limitation, we propose MDMR-Bench, a Multi-Dimensional benchmark for Multi-Reference image generation. MDMR-Bench is organized around representative capability settings and consists of multiple evaluation subsets spanning diverse dimensions of multi-reference image generation, including (1) common entity fusion and attribute transfer, (2) multi-element extraction from a single image, (3) infographic manipulation, and (4) visual reasoning-intensive generation. This design enables systematic analysis of model capabilities and failure modes across different challenge dimensions. Furthermore, to achieve more reliable and interpretable evaluation, we introduce a checklist-based evaluation protocol that decomposes generation quality into fine-grained evaluation aspects. Instead of relying solely on holistic scoring, our framework uses structured checklists composed of both global-level and atomic-level queries, enabling more diagnostic assessment of model performance. By benchmarking both open-source and proprietary models, we demonstrate that MDMR-Bench provides fine-grained insights into model performance across different dimensions. The dataset and code are publicly available.
https://arxiv.org/abs/2610.32459
Alzheimer's disease (AD) is a progressive neurological disorder associated with memory loss, cognitive decline, and behavioral disturbances. Early and accurate diagnosis is essential, yet MRI-based deep learning models often lack interpretability, limiting their clinical use. This study presents an explainable deep learning approach for four-class AD staging using 2D axial MRI images. The method combines eigenvector-based Grad-CAM with gradient smoothing to improve saliency localization and heatmap stability. Principal component analysis is applied to gradient-weighted convolutional layers, while Gaussian noise is added to saliency weights before decomposition. To address class imbalance in the Moderate Dementia category, Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) generates synthetic samples to balance the four-stage training distribution. Custom convolutional neural networks are trained on the Best Alzheimer MRI Dataset, achieving 97.2% test accuracy, 97.0% balanced accuracy, and a Matthews correlation coefficient of 0.944. The area under the ROC curve ranges from 0.96 to 0.99 across classes. Generated heatmaps highlight the hippocampus and entorhinal cortex, regions associated with Alzheimer's disease biomarkers. These findings suggest that the proposed approach combines strong classification performance with improved model interpretability, although further clinical validation is needed.
https://arxiv.org/abs/2610.32374
Controlling scientific instruments requires parameters to be adjusted progressively in response to feedback from successive operations. When unobserved operating conditions evolve with operational history, responses become nonstationary, and a static mapping from actions to responses becomes inadequate for decisions across shots. Shot-to-shot tuning of neutral beam injection (NBI) ion sources in magnetic confinement fusion devices is representative of this class of problems: the conditions of electrode surfaces and vacuum are difficult to observe directly, discharge outcomes vary between shots, and experiments are costly and carry operational risks. We propose a physically grounded multimodal agent with three independently auditable components. Deterministic algorithms extract symbolic physical quantities and visual evidence from multichannel waveforms; a memory with two layers organizes trajectories from the current task, experience from earlier tasks, and procedural knowledge, from which a multimodal large language model (MLLM) generates candidate actions; and an independent action feasibility check enforces action constraints before execution. In a strictly chronological replay of 455 shots from four EXL-50U platforms, physical symbols alone outperformed waveform images alone by 0.21-0.32 in macro-F1 for quality classification across all four evaluated models. Retrieved context increased the mean agreement with historical action directions by 22.4-42.7 percentage points, whereas the contribution of visual evidence varied across models. Under supervised operation on five ion sources, the system achieved a beam-extraction rate of 86.80% across 409 executed shots. These results demonstrate the engineering feasibility of the framework for supervised sequential tuning on a physical scientific instrument.
https://arxiv.org/abs/2610.32370
Music-driven group dance generation must decide not only how each dancer moves, but also what formation the group should form, when it should change, and who should execute each transition. Existing methods can generate orderly and rhythmically synchronized group motion, but remain limited in formation evolution and inter-dancer interaction. In human choreography, formation transitions often align with changes in high-level musical structure, such as the progression from an introduction to a chorus or from a chorus to a climax. We present ChoreoStudio, which makes choreography planning explicit. A music disentangler first decomposes the audio into frame-level rhythmic features and segment-level structural features. Given the segment-level features and sparse labels, a frozen omni-modal model acts as a scriptwriter, producing a global semantic description that guides the Director's choreography planning. Based on this context, the Director produces three coupled interaction decisions: a centroid-relative point set specifies what formation to produce, a transition gate determines when it should change, and a mobility allocation indicates who should execute the transition. The transition gate and mobility allocation are further conditioned on the planned geometric change, coupling formation targets with their temporal and dancer-specific execution. At a finer temporal scale, frame-aligned music features condition the diffusion generator to control detailed body motion and preserve beat-level alignment. The segment-level interaction plan is injected in parallel through frame-aligned instruction tokens and cross-dancer attention biases, coordinating how the generator realizes each transition. Experiments on AIOZ-GDance demonstrate that ChoreoStudio follows the planned interactions while achieving state-of-the-art performance in motion quality, group organization, and formation dynamics.
https://arxiv.org/abs/2610.32325
Point-supervised Temporal Action Localization (PTAL) aims to localize action instances using only a single annotated timestamp for each instance. Such sparse supervision offers limited guidance for modeling the structural organization of an action instance. Existing PTAL methods primarily infer action regions from classification responses, leaving the structural organization of multiple action-related parts within each instance insufficiently modeled. To address this problem, we propose DeformStruct, an instance-level deformable structure modeling framework for PTAL. DeformStruct progressively constructs the structure of each action instance through three stages. First, Structural Anchor Mining discovers multiple anchors from annotated points and action-responsive regions, providing structural nodes grounded in the current instance. Second, Deformable Part Alignment associates latent structural parts with these anchors through deformation-aware matching, allowing their temporal supports to adapt to instance-dependent variations. Third, Relational Structural Organization explicitly models the relative temporal relations among the aligned parts, forming a relational representation that characterizes how different action-related parts are organized together. Extensive experiments on THUMOS14, ActivityNet-v1.3, GTEA, and BEOID demonstrate consistent improvements in point-supervised temporal localization and validate the effectiveness of deformable structural modeling.
https://arxiv.org/abs/2610.32321
Video moment retrieval aims to localize a query-relevant temporal segment in an untrimmed video. Existing methods typically perform video-text alignment in Euclidean space, where representations are embedded in a flat geometry. However, accurate moment localization requires organizing query-relevant evidence across multiple grounding states, where target regions, boundary transitions, and contextual distractors exhibit distinct temporal structures and relevance patterns. Such Euclidean representations lack an explicit geometry to jointly capture these relationships, motivating us to explore structured temporal evidence organization from a geometric perspective. Specifically, we propose Hyperbolic Temporal Evidence Organization (HTEO), a geometric framework that organizes query-aware temporal memory in hyperbolic space. HTEO maps temporal representations into a Poincaré ball and organizes temporal evidence through Radial Evidence Organization (REO) and Orthogonal Angular Refinement (OAR). REO adjusts temporal tokens' radial positions to encode the structural differences among temporal evidence, while OAR enhances query-conditioned semantic alignment through orthogonal rotation. The organized representation is then projected back to the Euclidean tangent space and injected into the decoder through a lightweight residual memory update, enabling hyperbolic geometric cues to be incorporated into standard Transformer decoders while preserving the original cross-modal memory. Furthermore, we introduce radial structure regularization and angular relevance supervision to explicitly guide temporal evidence organization and query-conditioned semantic alignment, respectively. Extensive experiments and analyses across multiple benchmarks demonstrate that HTEO improves temporal localization while providing an interpretable geometric organization of temporal evidence.
https://arxiv.org/abs/2610.32317
LLM agents are increasingly used to solve real-world problems, and their effectiveness depends on both the underlying models and the harnesses that support their operation. However, most existing benchmarks focus on model capabilities and overall agent performance, with limited attention to the evaluation of agent harnesses. To address this gap, we introduce TraceTrial-Bench, which organizes harness reliability into seven components covering context, execution, governance, lifecycle, observability, tools, and verification. We construct 95 expert-level, mechanism-oriented tasks with executable or state-based ground truth to support semantically equivalent comparisons while preserving the native behavior of each harness. We design a Trajectory Tracing (TT) module to capture agent execution traces and introduce a two-layer evaluation framework that separately assesses agent performance and harness reliability. In addition, we design a Multi-source Evidence Acquisition (MEA) module to provide evidence from the agent context, harness, and execution environment for LLM-based evaluation, and develop a Failure Attribution (FA) module that uses explicit mechanism chains to support attribution at the run level. Across 4 harnesses, 9 models, and 95 tasks, our evaluation of 3,420 runs yields macro-averaged harness scores ranging from 59.8% to 72.0%, while 31.1% of runs achieve task success despite failing at least one harness criterion. TraceTrial-Bench provides an auditable and reproducible foundation for comparing agent harnesses, diagnosing their failures, and studying how harness design interacts with model behavior.
https://arxiv.org/abs/2610.32285
Traditional heuristic methods have accumulated mature experience in core objectives, such as space utilization and layout compactness of 3D bin packing, relying on artificially designed rule-based logic. However, they exhibit obvious limitations in generalization ability and adaptability to dynamic scenarios. Diffusion reinforcement learning, which integrates the spatial search capability of reinforcement learning and the data distribution modeling capability of diffusion learning, has achieved certain progress in action distribution modeling and decision compliance constraints. Nevertheless, there remain issues including insufficient perception of the spatial topological structure in 3D bin packing and limited layout optimization capability. To address these problems, we propose a diffusion reinforcement learning method integrated with spatial heuristic strategies, which combines classical bin packing heuristic rules with a diffusion reinforcement learning framework. Specifically, it compresses the action space through heuristic action pre-screening, introduces a graph neural network to model the spatial structure, and designs a composite reward function incorporating flatness to suppress spatial fragmentation, thereby realizing the organic integration of prior knowledge and adaptive optimization. Experimental results demonstrate that the pro
https://arxiv.org/abs/2610.32241
Affective computing is an interdisciplinary field that studies computational methods that relate to or influence emotion. These methods have been applied to interactive media artworks, but they have focused on affect detection rather than affect generation. For affect generation, computationally creative methods need to be explored that have recently been driven by the use of Generative Adversarial Networks (GANs), a deep learning method. The experiment presented in this paper, Forging Emotions, explores the use of visual emotion datasets and the working processes of GANs for visual affect generation, that is, for generating images that can convey or trigger specified emotions. This experiment concludes that the methodology used so far by computer science researchers to build image datasets for describing high-level concepts such as emotions is insufficient and proposes utilizing emotional networks of associations according to psychology research. Forging Emotions also concludes that to generate affect visually, merely corresponding to basic psychology findings, such as bright or dark colours, does not seem adequate. Therefore, research efforts should aim to understand the structure of trained GANs and compositional GANs in order to produce genuinely novel compositions that can convey or trigger emotions through the subject matter of generated images.
https://arxiv.org/abs/2610.32163
Neural networks can learn multiple tasks, but how learning organizes shared computation into specialized and reusable components remains poorly understood. In this work, we show that sparsity inducing regularization promotes functional modularity in multitask networks without predefined module boundaries. Using functional profiling in multilayer perceptrons and a lightweight transformer, we identified compact, task-selective subnetworks, termed functional tokens. Across rule-based task families, selectively retaining these tokens preserves their associated functions while reducing performance on other tasks. Such specialization develops preferentially in deeper layers during training, while earlier layers retain shared computation. Regularization strength and form shape module compactness and specialization, while task weighting alters functional allocation. Moreover, extracted functional tokens can be recombined through alternative pathways to reconstruct a learned composite computation without retraining, preserving their connectivity and learned weights. Our results suggest that sparsity-constrained optimization shapes not only parameter use but also the organization and reuse of learned computation.
https://arxiv.org/abs/2610.32153
Proactive streaming video understanding requires assistants to track unfolding events and respond when a user's requested condition is met. To make these decisions, assistants must accumulate relevant evidence and reassess response conditions as new observations arrive. Existing work offers complementary capabilities for this process: proactive video assistants learn when to respond, while streaming video reasoning methods generate scheduled intermediate text to organize observations and support subsequent answers. Together, these capabilities provide a foundation for proactive reasoning, but how to coordinate ongoing reasoning with timely responses remains underexplored. In a direct adaptation, fixed reasoning intervals can trigger updates when nothing relevant has changed or postpone useful updates until the next scheduled step. Sharing one generation path also couples two distinct requirements: preserving evidence for future decisions and producing only the response requested by the user. Moreover, responses generated after reasoning must wait for it to finish, even when sufficient evidence is already available. To address these challenges, we introduce SyncPro, a synchronous dual-stream framework with specialized generation paths for evidence tracking and user-facing responses. Its Inner and Response branches share causal history and activate independently, allowing relevant evidence changes to trigger internal updates and responses to proceed concurrently with ongoing reasoning. We construct 189K dual-stream examples and train SyncPro through supervised fine-tuning followed by trajectory-level reinforcement learning. Experiments across three proactive benchmarks show improved response timing and more accurate answers to user requests.
https://arxiv.org/abs/2610.32132
Understanding language-agent behaviour requires learning how actions are organised across a trajectory and how this organisation relates to context and outcomes. We propose a general framework that connects data-driven behavioural discovery with analysis and experimental investigation. Categories developed from recorded activity define what is measured. Analyses preserve the behaviours' positions in the process and express their relationships as interpretable hypotheses. Statistical assessment and tests in held-out contexts examine the strength and scope of the evidence. Experiments, including prompt steering, then test how behaviour and outcomes change under deliberately varied conditions. A study of iterative self-reflection illustrates how discovery, dependency learning and prompt comparisons can inform one another. We develop this approach for trajectories of varying length, event order and agent roles, and identify advances in coding, analysis and experimentation needed to support its wider use.
https://arxiv.org/abs/2610.32087
The rapid growth of microdramas has increased demand for high-quality script production. Long-form microdrama script generation with LLMs requires both diverse plot progression and long-range narrative coherence. Our observations reveal that candidate continuations increasingly concentrate on fewer narrative mechanisms as generation progresses, while conventional retrieval may reinforce trope reuse and overlook historical information essential to future plots. To address these challenges, we propose DraMaze, a multi-agent pipeline that explores diverse causal paths through trope memory-guided evolutionary tree search and mutation. It further constructs a Backward Narrative Dependency Tree to organize, retain, and retrieve historical memories based on future narrative goals. We also introduce MiDramaBench, a bilingual benchmark for evaluating script logic and creative quality across different script lengths, together with causal-pattern metrics for narrative repetition and diversity. Experiments show consistent improvements across three backbone LLMs. For 60-episode scripts of up to 60k words, DraMaze reduces causal-pattern repetition by 32.1% and achieves a Logic score of 82.5, outperforming the best-performing baseline by 8.5 points.
https://arxiv.org/abs/2610.32002
The cardiac attractor provides a phase-space representation of cardiovascular dynamics reconstructed from scalar observations under standard delay-embedding conditions. We introduce Attractor Domain Theory (ADT), a theory in which linked mathematical transformations expose distinct properties of reconstructed dynamics: the Geometry Domain is obtained through the delay-reconstruction transformation, while Ergodic and Variational transformations act on that geometry to characterize long-horizon organization and localized finite-time deformation. Thus, ADT specifies a three-representation operator architecture for reconstructed dynamics rather than retrospectively grouping existing nonlinear features; the architecture is proposed as an operational organization, not as a uniqueness or exhaustiveness theorem. A Conditional Domain Decomposition quantifies endpoint-specific incremental information with an explicit residual, and a domain-transformation impossibility theorem establishes that a property cannot be exactly recovered when an operator maps property-distinct inputs to the same representation. Empirical evaluation includes PPG analysis across 176{,}742 segments with 18 patient-level held-out BIDMC records (AUC 0.757 [0.686--0.828]) and an independent ECG analysis of 5{,}000 PTB-XL records (AUC 0.719 [0.686--0.750]). Cross-observation transfer is property-specific: the Lyapunov exponent shows subject-level ECG--PPG agreement ($\rho=0.703$, 95\% CI [0.503, 0.836]), whereas the predicted class-wide separation of invariant and observation-coupled descriptors is not supported ($p=0.54$). These results provide empirical support for attractor-domain analysis across cardiovascular measurements while leaving cross-observation recoverability as a property-specific empirical question.
https://arxiv.org/abs/2606.22039
Vision language models (VLMs) have been increasingly adopted in consumer smart-home camera systems to interpret recorded videos and provide insights. However, video clips captured from smart-home cameras are different from curated web videos dominating existing VLM benchmarks. Smart-home video clips often do not contain subtitles or narration and are captured from different viewing angles, under various lighting conditions, and in diverse scenarios. However, the capabilities of the model in this domain, under sparse visual signals and limited context, are not fully assessed by existing VLM benchmarks. To close this gap, we introduce OpenHome, a visual-only, four-option multiple-choice benchmark for evaluating smart-home video understanding. It contains 2,254 questions on 1,118 real-world video clips and organizes 10 tasks into three capability levels. Perception evaluates visible content, Reasoning evaluates causal, spatial, and temporal relationships, and Insight evaluates subject intent, supporting evidence and appropriate homeowner responses. Based on the dataset, we describe the task design, data pipeline and evaluation protocol, and report results for 32 closed-source and open-weight VLMs. The best model, Gemini 4 Argon, reaches 81.2% overall accuracy. Frontier closed-source models outperform open-weight models on every level. The best closed-source model leads the best open-weight model by 16.3% on Perception, 22.4% on Reasoning and 37.4% on Insight.
https://arxiv.org/abs/2610.31986
Agentic AI systems using the Model Context Protocol (MCP) are vulnerable to indirect prompt injection through untrusted tool outputs. We demonstrate cross-modal trust laundering, in which instructions concealed in an email-signature image using least-significant-bit (LSB), discrete cosine transform (DCT), or metadata steganography bypass text scanners and enter a multi-agent workflow when decoded by a tool. To counter this threat, we propose decode-time provenance binding, which combines HMAC-authenticated trust labels with an MCP-maintained provenance ledger and context-taint tracking. The mechanism labels decoded content as untrusted and blocks critical tool calls influenced by that content, even when a compromised agent substitutes a valid trusted label. We evaluate the approach using a real MCP testbed with three steganographic carriers and 24 adaptive payloads. The undefended pipeline exhibits 100% attack success, while keyword and regex defenses each permit 62.5% of attacks. Our mechanism blocks all attacks (0% success), including adversarial label substitution across mediated agent boundaries. A four-component ablation validates the roles of decode marking, cryptographic authentication, lineage tracking, and sink enforcement. An audited trusted-promotion mechanism increases legitimate-task completion from 67% to 100% without compromising protection. The approach provides carrier-independent containment within fully mediated MCP execution environments. Code and data are publicly available.
https://arxiv.org/abs/2610.31974
Large language model (LLM) agents increasingly participate in negotiations on behalf of individuals and organizations, yet they remain vulnerable to manipulation, particularly in repeated multi-party interactions. Classical multi-agent systems address this challenge through explicit trust and reputation models, whereas existing work on LLM negotiation largely treats trust as an emergent phenomenon. We investigate whether LLM agents can form useful trust estimates from prior interactions and leverage them to improve future negotiation outcomes. We introduce a repeated, multi-party, multi-issue negotiation framework in which recurring agents participate in a sequence of negotiation games while one participant covertly attempts to prevent agreement. After each game, agents assess their counterparts and maintain pairwise trust estimates that carry over to subsequent negotiations. We compare trust-aware and trust-agnostic agents and analyze how explicit trust influences negotiation outcomes. Our results reveal a \emph{trust--agreement gap}: trust-aware agents can identify the adversary within a few games, but accurate trust estimates do not consistently translate into better agreements. Trust increases agreement rates by up to 23 percentage points, but only when agents are explicitly guided to use it in their decisions. Explicit trust can therefore improve repeated LLM negotiations, but its value depends critically on how trust is assessed and incorporated into agents' reasoning.
https://arxiv.org/abs/2610.31971
Physical self-reproduction in modular robots remains difficult because a robot must not only move, but also collect, align, dock, and separate modules reliably in hardware. Taking inspiration from how genetically identical cells form the varied tissues of an organism, we introduce a single standardized robotic module that combines a magnetic docking interface with two actuated joints and folds from a flat, two-dimensional configuration into three-dimensional structures. Modules of this single design can be assembled into two morphologies: a three-module "trilobite" and a five-module "seal". Using a small set of discrete motion commands, a human operator steers an assembled parent robot to dock magnetically with a sequence of spare modules, which are assembled into a second, structurally identical individual that then separates from the parent. Because the parent retains all of its original modules, this is a budding-style form of physical reproduction. Here, we report preliminary hardware demonstrations of this full process on both morphologies. We discuss the current scope of the system, including its present reliance on direct human teleoperation rather than autonomous control, and outline directions toward autonomous execution of this process and repeated validation across a larger number of independent trials.
https://arxiv.org/abs/2610.31929
Recent Sparse Autoencoders (SAEs) decompose the representations of large language models (LLMs) into interpretable features, enabling precise readout and control of encoded knowledge. However, existing SAEs treat each feature as a standalone unit, whereas real-world knowledge is often taxonomically organized. As a result, readout carries no guarantee of taxonomic consistency, and control lacks the granularity to specify the scope of an edit. We propose TaxoSAE, which explicitly encodes taxonomical information, built on is-a (hypernym-hyponym) relations, into the SAE feature space. TaxoSAE mirrors a taxonomy with a stack of SAEs, one per taxonomy level, and introduces a taxonomic alignment loss that trains each concept to be active whenever one of its hyponyms is. On two factual QA taxonomies, one over biographical relations and one over animals, TaxoSAE enables (1) more accurate readout through the taxonomy with concentrated activations, e.g., +31.3 points in reading out the hypernyms of abstract concepts never queried in training; (2) more precise control, with +25.5 points in swap success, while suppressing a concept removes only its own answers (0-2% accuracy vs. 91-100% on other concepts); and, more importantly, (3) readout of unseen concepts through their hypernyms, with up to +37.9 points in hypernym readout accuracy.
https://arxiv.org/abs/2610.31894