# Research Papers — 2026-04 · bundle page 203 of 233 > Full-text markdown of papers #10101–10150 of 11612 published in 2026-04, newest first. AI-readable mirrors of the full 11612-paper 2026-04 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-04/page/203 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/202.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/204.md) ### RCP: Representation Consistency Pruner for Mitigating Distribution Shift in Large Vision-Language Models Authors: Jianwei Zhang, Chaoning Zhang, Sihan Cao, Wang Liu, Pengcheng Zheng, Jiaxin Huang, Caiyan Qin, Yalan Ye, Wei Dong, Yang Yang Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.04972v1 Large Vision-Language Models (LVLMs) suffer from prohibitive inference costs due to the massive number of visual tokens processed by the language decoder. Existing pruning methods often lead to significant performance degradation because the irreversible removal of visual tokens causes a distribution shift in the hidden states that deviates from the pre-trained full-token regime. To address this, we propose Representation Consistency Pruner, which we refer to as RCP, as a novel framework that integrates cumulative visual token pruning with a delayed repair mechanism. Specifically, we introduce a cross-attention pruner that leverages the intrinsic attention of the LLM as a baseline to predict cumulative masks, ensuring consistent and monotonic token reduction across layers. To compensate for the resulting information loss, we design a delayed repair adapter denoted as DRA, which caches the essence of pruned tokens and applies FiLM-based modulation specifically to the answer generation tokens. We employ a repair loss to match the first and second-order statistics of the pruned representations with a full-token teacher. RCP is highly efficient because it trains only lightweight plug-in modules while allowing for physical token discarding at inference. Extensive experiments on LVLM benchmarks demonstrate that RCP removes up to 88.9\% of visual tokens and reduces FLOPs by up to 85.7\% with only a marginal average accuracy drop, and outperforms prior methods that avoid fine-tuning the original model on several widely used benchmarks. --- ### A Theory-guided Weighted $L^2$ Loss for solving the BGK model via Physics-informed neural networks Authors: Gyounghun Ko, Sung-Jun Son, Seung Yeon Cho, Myeong-Su Lee Published: 2026-04-04 Categories: cs.LG, math.NA, physics.comp-ph arXiv: https://arxiv.org/abs/2604.04971v1 While Physics-Informed Neural Networks offer a promising framework for solving partial differential equations, the standard $L^2$ loss formulation is fundamentally insufficient when applied to the Bhatnagar-Gross-Krook (BGK) model. Specifically, simply minimizing the standard loss does not guarantee accurate predictions of the macroscopic moments, causing the approximate solutions to fail in capturing the true physical solution. To overcome this limitation, we introduce a velocity-weighted $L^2$ loss function designed to effectively penalize errors in the high-velocity regions. By establishing a stability estimate for the proposed approach, we shows that minimizing the proposed weighted loss guarantees the convergence of the approximate solution. Also, numerical experiments demonstrate that employing this weighted PINN loss leads to superior accuracy and robustness across various benchmarks compared to the standard approach. --- ### MG$^2$-RAG: Multi-Granularity Graph for Multimodal Retrieval-Augmented Generation Authors: Sijun Dai, Qiang Huang, Xiaoxing You, Jun Yu Published: 2026-04-04 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2604.04969v2 Retrieval-Augmented Generation (RAG) mitigates hallucinations in Multimodal Large Language Models (MLLMs), yet existing systems struggle with complex cross-modal reasoning. Flat vector retrieval often ignores structural dependencies, while current graph-based methods rely on costly ``translation-to-text'' pipelines that discard fine-grained visual information. To address these limitations, we propose \textbf{MG$^2$-RAG}, a lightweight \textbf{M}ulti-\textbf{G}ranularity \textbf{G}raph \textbf{RAG} framework that jointly improves graph construction, modality fusion, and cross-modal retrieval. MG$^2$-RAG constructs a hierarchical multimodal knowledge graph by combining lightweight textual parsing with entity-driven visual grounding, enabling textual entities and visual regions to be fused into unified multimodal nodes that preserve atomic evidence. Building on this representation, we introduce a multi-granularity graph retrieval mechanism that aggregates dense similarities and propagates relevance across the graph to support structured multi-hop reasoning. Extensive experiments across four representative multimodal tasks (i.e., retrieval, knowledge-based VQA, reasoning, and classification) demonstrate that MG$^2$-RAG consistently achieves state-of-the-art performance while reducing graph construction overhead with an average 43.3$\times$ speedup and 23.9$\times$ cost reduction compared with advanced graph-based frameworks. --- ### Belief Dynamics for Detecting Behavioral Shifts in Safe Collaborative Manipulation Authors: Devashri Naik, Divake Kumar, Nastaran Darabi, Amit Ranjan Trivedi Published: 2026-04-04 Categories: cs.RO, cs.LG arXiv: https://arxiv.org/abs/2604.04967v1 Robots operating in shared workspaces must maintain safe coordination with other agents whose behavior may change during task execution. When a collaborating agent switches strategy mid-episode, continuing under outdated assumptions can lead to unsafe actions and increased collision risk. Reliable detection of such behavioral regime changes is therefore critical. We study regime-switch detection under controlled non-stationarity in ManiSkill shared-workspace manipulation tasks. Across ten detection methods and five random seeds, enabling detection reduces post-switch collisions by 52%. However, average performance hides significant reliability differences: under a realistic tolerance of +-3 steps, detection ranges from 86% to 30%, while under +-5 steps all methods achieve 100%. We introduce UA-TOM, a lightweight belief-tracking module that augments frozen vision-language-action (VLA) control backbones using selective state-space dynamics, causal attention, and prediction-error signals. Across five seeds and 1200 episodes, UA-TOM achieves the highest detection rate among unassisted methods (85.7% at +-3) and the lowest close-range time (4.8 steps), outperforming an Oracle (5.3 steps). Analysis shows hidden-state update magnitude increases by 17x at regime switches and decays over roughly 10 timesteps, while the discretization step converges to a near-constant value (Delta_t approx 0.78), indicating sensitivity driven by learned dynamics rather than input-dependent gating. Cross-domain experiments in Overcooked show complementary roles of causal attention and prediction-error signals. UA-TOM introduces 7.4 ms inference overhead (14.8% of a 50 ms control budget), enabling reliable regime-switch detection without modifying the base policy. --- ### LLM-Agent-based Social Simulation for Attitude Diffusion Authors: Deepak John Reji Published: 2026-04-04 Categories: cs.AI, stat.CO arXiv: https://arxiv.org/abs/2604.03898v1 This paper introduces discourse_simulator, an open-source framework that combines LLMs with agent-based modelling. It offers a new way to simulate how public attitudes toward immigration change over time in response to salient events like protests, controversies, or policy debates. Large language models (LLMs) are used to generate social media posts, interpret opinions, and model how ideas spread through social networks. Unlike traditional agent-based models that rely on fixed, rule-based opinion updates and cannot generate natural language or consider current events, this approach integrates multidimensional sociological belief structures and real-world event timelines. This framework is wrapped into an open-source Python package that integrates generative agents into a small-world network topology and a live news retrieval system. discourse_sim is purpose-built as a social science research instrument specifically for studying attitude dynamics, polarisation, and belief evolution following real-world critical events. Unlike other LLM Agent Swarm frameworks, which treat the simulations as a prediction black box, discourse_sim treats it as a theory-testing instrument, which is fundamentally a different epistemological stance for studying social science problems. The paper further demonstrates the framework by modelling the Dublin anti-immigration march on April 26, 2025, with N=100 agents over a 15-day simulation. Package link: https://pypi.org/project/discourse-sim/ --- ### Latency-Aware Resource Allocation over Heterogeneous Networks: A Lorentz-Invariant Market Mechanism Authors: Saad Alqithami Published: 2026-04-04 Categories: cs.GT, cs.AI, cs.NI arXiv: https://arxiv.org/abs/2604.03897v1 We present a telecom-native auction mechanism for allocating bandwidth and time slots across heterogeneous-delay networks, ranging from low-Earth-orbit (LEO) satellite constellations to delay-tolerant deep-space relays. The Lorentz-Invariant Auction (LIA) treats bids as spacetime events and reweights reported values based on the \emph{horizon slack}, a causal quantity derived from the earliest-arrival times relative to a public clearing horizon. Unlike other delay-equalization rules, LIA combines a causal-ordering formulation, a uniquely exponential slack correction implied by a semigroup-style invariance axiom, and a critical-value implementation that ensures truthful reported values once slacks are fixed by trusted infrastructure. We analyze the incentive result in the exogenous-slack regime and separately examine bounded slack-estimation error and endogenous-delay limitations. Under fixed feasible slacks, LIA is individually rational and achieves welfare at least \(e^{-λΔ}\) relative to the optimal feasible allocation, where \(Δ\) is the slack spread. We evaluate LIA on STARLINK-200, INTERNET-100, and DSN-30 across 52,500 baseline instances with market sizes \(n\in\{10,20,30,40,50\}\) and conduct additional robustness sweeps. On Starlink and Internet, LIA maintains near-efficiency while eliminating measured timing rents. However, on DSN, welfare is lower in thin markets but improves with depth. We also distinguish winner-determination time from the background cost of maintaining slack estimates and study robustness beyond independent and identically distributed (iid) noise through error-spread bounds and structured (distance-biased and subnetwork-correlated) noise models. These results suggest that causal-consistent mechanism design offers a practical non-buffering alternative to synchronized delay equalization in heterogeneous telecom infrastructures. --- ### FeynmanBench: Benchmarking Multimodal LLMs on Diagrammatic Physics Reasoning Authors: Zeyu Wang, Jingye Xu, Xiaogang Li, Peiyao Xiao, Qinhao Kong, Ben Wang, Chengliang Xu, Zichao Chen, Bing Zhao, Hu Wei Published: 2026-04-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.03893v2 Current multimodal benchmarks for scientific reasoning primarily evaluate local information extraction -- models recognize symbols and values and then perform textual inference. They do not assess whether models can reason over the global structural properties of formal diagrams, such as topology, conservation constraints, and the consistent mapping between visual patterns and algebraic expressions. We introduce FeynmanBench, a benchmark of over 2,000 tasks centered on Feynman diagrams spanning the electromagnetic, weak, and strong interactions of the Standard Model. Each instance couples a diagram image with minimal textual conventions and requires models to recover the full physical content -- vertex inventory, propagator types, topological connectivity, momentum routing, and the complete scattering amplitude. An automated generation and verification pipeline produces the diagrams, annotations, and reference answers under standardized rules. Evaluating 19 state-of-the-art multimodal LLMs, we find a consistent failure pattern: models achieve 70--95\% on local recognition (vertex and propagator identification) but collapse to 13--17\% on topological reconstruction (CP3), and near zero on full algebraic derivation (CP5). FeynmanBench offers a controlled testbed for multimodal reasoning over formal scientific diagrams and highlights fundamental limitations of current architectures in topology-sensitive scientific reasoning. --- ### Lotka-Sharpe Neural Operators for Control of Population PDEs Authors: Miroslav Krstic, Iasson Karafyllis, Luke Bhan, Carina Veil Published: 2026-04-04 Categories: eess.SY, cs.LG, math.OC arXiv: https://arxiv.org/abs/2604.03892v1 Age-structured predator-prey integro-partial differential equations provide models of interacting populations in ecology, epidemiology, and biotechnology. A key challenge in feedback design for these systems is the scalar $ζ$, defined implicitly by the Lotka-Sharpe nonlinear integral condition, as a mapping from fertility and mortality rates to $ζ$. To solve this challenge with operator learning, we first prove that the Lotka-Sharpe operator is Lipschitz continuous, guaranteeing the existence of arbitrarily accurate neural operator approximations over a compact set of fertility and mortality functions. We then show that the resulting approximate feedback law preserves semi-global practical asymptotic stability under propagation of the operator approximation error through various other nonlinear operators, all the way through to the control input. In the numerical results, not only do we learn ``once-and-for-all'' the canonical Lotka-Sharpe (LS) operator, and thus make it available for future uses in control of other age-structured population interconnections, but we demonstrate the online usage of the neural LS operator under estimation of the fertility and mortality functions. --- ### Provable Multi-Task Reinforcement Learning: A Representation Learning Framework with Low Rank Rewards Authors: Yaoze Guo, Shana Moothedath Published: 2026-04-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.03891v1 Multi-task representation learning (MTRL) is an approach that learns shared latent representations across related tasks, facilitating collaborative learning that improves the overall learning efficiency. This paper studies MTRL for multi-task reinforcement learning (RL), where multiple tasks have the same state-action space and transition probabilities, but different rewards. We consider T linear Markov Decision Processes (MDPs) where the reward functions and transition dynamics admit linear feature embeddings of dimension d. The relatedness among the tasks is captured by a low-rank structure on the reward matrices. Learning shared representations across multiple RL tasks is challenging due to the complex and policy-dependent nature of data that leads to a temporal progression of error. Our approach adopts a reward-free reinforcement learning framework to first learn a data-collection policy. This policy then informs an exploration strategy for estimating the unknown reward matrices. Importantly, the data collected under this well-designed policy enable accurate estimation, which ultimately supports the learning of an near-optimal policy. Unlike existing approaches that rely on restrictive assumptions such as Gaussian features, incoherence conditions, or access to optimal solutions, we propose a low-rank matrix estimation method that operates under more general feature distributions encountered in RL settings. Theoretical analysis establishes that accurate low-rank matrix recovery is achievable under these relaxed assumptions, and we characterize the relationship between representation error and sample complexity. Leveraging the learned representation, we construct near-optimal policies and prove a regret bound. Experimental results demonstrate that our method effectively learns robust shared representations and task dynamics from finite data. --- ### PolySwarm: A Multi-Agent Large Language Model Framework for Prediction Market Trading and Latency Arbitrage Authors: Rajat M. Barot, Arjun S. Borkhatariya Published: 2026-04-04 Categories: cs.AI, cs.CL, cs.MA, q-fin.TR arXiv: https://arxiv.org/abs/2604.03888v1 This paper presents PolySwarm, a novel multi-agent large language model (LLM) framework designed for real-time prediction market trading and latency arbitrage on decentralized platforms such as Polymarket. PolySwarm deploys a swarm of 50 diverse LLM personas that concurrently evaluate binary outcome markets, aggregating individual probability estimates through confidence-weighted Bayesian combination of swarm consensus with market-implied probabilities, and applying quarter-Kelly position sizing for risk-controlled execution. The system incorporates an information-theoretic market analysis engine using Kullback-Leibler (KL) divergence and Jensen-Shannon (JS) divergence to detect cross-market inefficiencies and negation pair mispricings. A latency arbitrage module exploits stale Polymarket prices by deriving CEX-implied probabilities from a log-normal pricing model and executing trades within the human reaction-time window. We provide a full architectural description, implementation details, and evaluation methodology using Brier scores, calibration analysis, and log-loss metrics benchmarked against human superforecaster performance. We further discuss open challenges including hallucination in agent pools, computational cost at scale, regulatory exposure, and feedback-loop risk, and outline five priority directions for future research. Experimental results demonstrate that swarm aggregation consistently outperforms single-model baselines in probability calibration on Polymarket prediction tasks. --- ### PhaseFlow4D: Physically Constrained 4D Beam Reconstruction via Feedback-Guided Latent Diffusion Authors: Alexander Scheinker, Alexander Plastun, Peter Ostroumov Published: 2026-04-04 Categories: physics.acc-ph, cs.LG arXiv: https://arxiv.org/abs/2604.03885v2 We address the problem of recovering a time-varying 4D distribution from a sparse sequence of 2D projections - analogous to novel-view synthesis from sparse cameras, but applied to the 4D transverse phase space density $ρ(x,p_x,y,p_y)$ of charged particle beams. Direct single shot measurement of this high-dimensional distribution is physically impossible in real particle accelerator systems; only limited 1D or 2D projections are accessible. We propose PhaseFlow4D, a feedback-guided latent diffusion model that reconstructs and tracks the full 4D phase space from incomplete 2D observations alone, with built-in hard physics constraints. Our core technical contribution is a 4D VAE whose decoder generates the full 4D phase space tensor, from which 2D projections are analytically computed and compared against 2D beam measurements. This projection-consistency constraint guarantees physical correctness by construction - not as a soft penalty, but as an architectural prior. An adaptive feedback loop then continuously tunes the conditioning vector of the latent diffusion model to track time-varying distributions online without retraining. We validate on multi-particle simulations of heavy-ion beams at the Facility for Rare Isotope Beams (FRIB), where full physics simulations require $\sim$6 hours on a 100-core HPC system. PhaseFlow4D achieves accurate 4D reconstructions 11000$\times$ faster while faithfully tracking distribution shifts under time-varying source conditions - demonstrating that principled generative reconstruction under incomplete observations transfers robustly beyond visual domains. --- ### Regime-Calibrated Fleet Repositioning with a Spatial Queue-Regret Decomposition Authors: Indar Kumar, Akanksha Tiwari Published: 2026-04-04 Categories: cs.LG, cs.AI, eess.SY, stat.ML arXiv: https://arxiv.org/abs/2604.03883v2 Ride-hailing and autonomous mobility-on-demand operators reposition idle supply before future demand is fully observed. We study a retrieval-calibrated predict-then-optimize approach for this problem: historical demand regimes are matched to the current query block, combined into a calibrated demand prior, and passed to a fleet-balancing controller. The paper makes three contributions. First, we train a leakage-safe similarity gate whose objective penalizes demand error, pickup spatial mismatch, and queue shortage risk rather than retrieval rank alone. Second, we develop a spatial queue-regret decomposition for a stable queueing surrogate, linking demand-field error to wait through queueing sensitivity, allocator sensitivity, and Wasserstein pickup mismatch. Third, we evaluate learned retrieval and external-style rebalancing baselines in a common simulator. In the calibrated-demand gate experiment, across eight New York City scenarios and ten seeds, the spatial gate reduces mean wait to 82.3s, compared with 85.3s for hand-tuned similarity and 85.8s for a distributional-only baseline. In a separate replay-demand controller comparison, a scenario chance-MPC analog and a share-target transportation LP improve on Wen-style rebalancing (92.2s/92.2s vs. 100.1s), a reduced GPR chance-MPC comparator is intermediate at 94.4s, and an oracle MPC diagnostic is 91.3s. --- ### LLM-generated personalized nudges for improving pro-environmental behavior: Field evidence from resource conservation Authors: Zonghan Li, Yi Liu, Chunyan Wang, Song Tong, Kaiping Peng, Feng Ji Published: 2026-04-04 Categories: cs.CY, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2604.03881v2 Encouraging pro-environmental behavior remains a major challenge for sustainable cities. Conventional feedback nudges can show individuals how their current behavior compares with environmental goals but often provide limited guidance on what to do differently in daily life. This study examines whether supplementing weekly feedback on participants' behavior with LLM-generated personalized action suggestions improves pro-environmental behavior, using daily electricity and hot-water conservation as a case study. We developed an LLM agent that generated weekly conservation messages from participant profiles, recent consumption records, and prior interaction history, combining a usage report with personalized suggestions, behavioral-change scenarios, and estimated savings. The agent was evaluated in a three-arm randomized field experiment with 233 university residents in Beijing from November 2024 to January 2025. Participants received text-based nudges, image-enhanced nudges, or LLM-generated personalized nudges over five intervention rounds. Daily electricity use and shower hot-water use were measured using dormitory meter readings and billing records. Compared with text-based feedback, LLM-generated personalized nudges reduced electricity consumption by 0.56 kWh per room-day (p = 0.014), corresponding to an 18.3 percentage-point higher saving rate. Image-enhanced feedback alone showed no clear improvement. Hot-water savings followed the same direction but were smaller and less precisely estimated (9.8 percentage points, p = 0.087). Personalized nudges contained more planning, appliance-specific, and action-oriented language and were associated with more sustained, task-focused engagement. These findings offer a pathway for integrating generative AI into sustainable urban management. --- ### Learning 3D Reconstruction with Priors in Test Time Authors: Lei Zhou, Haoyu Wu, Akshat Dave, Dimitris Samaras Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03878v1 We introduce a test-time framework for multiview Transformers (MVTs) that incorporates priors (e.g., camera poses, intrinsics, and depth) to improve 3D tasks without retraining or modifying pre-trained image-only networks. Rather than feeding priors into the architecture, we cast them as constraints on the predictions and optimize the network at inference time. The optimization loss consists of a self-supervised objective and prior penalty terms. The self-supervised objective captures the compatibility among multi-view predictions and is implemented using photometric or geometric loss between renderings from other views and each view itself. Any available priors are converted into penalty terms on the corresponding output modalities. Across a series of 3D vision benchmarks, including point map estimation and camera pose estimation, our method consistently improves performance over base MVTs by a large margin. On the ETH3D, 7-Scenes, and NRGBD datasets, our method reduces the point-map distance error by more than half compared with the base image-only models. Our method also outperforms retrained prior-aware feed-forward methods, demonstrating the effectiveness of our test-time constrained optimization (TCO) framework for incorporating priors into 3D vision tasks. --- ### When Models Know More Than They Say: Probing Analogical Reasoning in LLMs Authors: Hope McGovern, Caroline Craig, Thomas Lippincott, Hale Sirin Published: 2026-04-04 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.03877v1 Analogical reasoning is a core cognitive faculty essential for narrative understanding. While LLMs perform well when surface and structural cues align, they struggle in cases where an analogy is not apparent on the surface but requires latent information, suggesting limitations in abstraction and generalisation. In this paper we compare a model's probed representations with its prompted performance at detecting narrative analogies, revealing an asymmetry: for rhetorical analogies, probing significantly outperforms prompting in open-source models, while for narrative analogies, they achieve a similar (low) performance. This suggests that the relationship between internal representations and prompted behavior is task-dependent and may reflect limitations in how prompting accesses available information. --- ### Neural Processes Maintain Calibrated Biomass Estimates Across Spatiotemporal Gaps and Disturbance Authors: Robin Young, Srinivasan Keshav Published: 2026-04-04 Categories: cs.LG, cs.CE arXiv: https://arxiv.org/abs/2604.03874v2 Monitoring deforestation-driven carbon emissions requires both spatially explicit and temporally continuous estimates of aboveground biomass density (AGBD) with calibrated uncertainty. NASA's Global Ecosystem Dynamics Investigation (GEDI) provides reliable LIDAR-derived AGBD, but its orbital sampling causes irregular spatiotemporal coverage, and occasional operational interruptions, including a 13-month hibernation from March 2023 to April 2024, leave extended gaps in the observational record. Prior work has used machine learning approaches to fill GEDI's spatial gaps using satellite-derived features, but temporal interpolation of biomass through unobserved periods, particularly across active disturbance events, remains largely unaddressed. Moreover, standard ensemble methods for biomass mapping have been shown to produce systematically miscalibrated prediction intervals. To address these gaps, we extend the Attentive Neural Process (ANP) framework, previously applied to spatial biomass interpolation, to jointly sparse spatiotemporal settings using geospatial foundation model embeddings. We treat space and time symmetrically, empirically validating a form of space-for-time substitution in which observations from nearby locations at other times inform predictions at held-out periods. Our results demonstrate that the ANP produces well-calibrated uncertainty estimates across disturbance regimes, supporting its use in Measurement, Reporting, and Verification (MRV) applications that require reliable uncertainty quantification for forest carbon accounting. --- ### SODA: Semi On-Policy Black-Box Distillation for Large Language Models Authors: Xiwen Chen, Jingjing Wang, Wenhui Zhu, Peijie Qiu, Xuanzhao Dong, Yueyue Deng, Hejian Sang, Zhipeng Wang, Alborz Geramifard, Feng Luo Published: 2026-04-04 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2604.03873v5 Black-box knowledge distillation for large language models presents a strict trade-off. Simple off-policy methods (e.g., sequence-level knowledge distillation) struggle to correct the student's inherent errors. Fully on-policy methods (e.g., Generative Adversarial Distillation) solve this via adversarial training but introduce well-known training instability and crippling computational overhead. To address this dilemma, we propose SODA (Semi On-policy Distillation with Alignment), a highly efficient alternative motivated by the inherent capability gap between frontier teachers and much smaller base models. Because a compact student model's natural, zero-shot responses are almost strictly inferior to the powerful teacher's targets, we can construct a highly effective contrastive signal simply by pairing the teacher's optimal response with a one-time static snapshot of the student's outputs. This demonstrates that exposing the small student to its own static inferior behaviors is sufficient for high-quality distribution alignment, eliminating the need for costly dynamic rollouts and fragile adversarial balancing. Extensive evaluations across four compact Qwen2.5 and Llama-3 models validate this semi on-policy paradigm. SODA matches or outperforms the state-of-the-art methods on 15 out of 16 benchmark results. More importantly, it achieves this superior distillation quality while training 10 times faster, consuming 27% less peak GPU memory, and completely eliminating adversarial instability. --- ### Your Agent is More Brittle Than You Think: Uncovering Indirect Injection Vulnerabilities in Agentic LLMs Authors: Wenhui Zhu, Xuanzhao Dong, Xiwen Chen, Rui Cai, Peijie Qiu, Zhipeng Wang, Oana Frunza, Shao Tang, Jindong Gu, Yalin Wang Published: 2026-04-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.03870v1 The rapid deployment of open-source frameworks has significantly advanced the development of modern multi-agent systems. However, expanded action spaces, including uncontrolled privilege exposure and hidden inter-system interactions, pose severe security challenges. Specifically, Indirect Prompt Injections (IPI), which conceal malicious instructions within third-party content, can trigger unauthorized actions such as data exfiltration during normal operations. While current security evaluations predominantly rely on isolated single-turn benchmarks, the systemic vulnerabilities of these agents within complex dynamic environments remain critically underexplored. To bridge this gap, we systematically evaluate six defense strategies against four sophisticated IPI attack vectors across nine LLM backbones. Crucially, we conduct our evaluation entirely within dynamic multi-step tool-calling environments to capture the true attack surface of modern autonomous agents. Moving beyond binary success rates, our multidimensional analysis reveals a pronounced fragility. Advanced injections successfully bypass nearly all baseline defenses, and some surface-level mitigations even produce counterproductive side effects. Furthermore, while agents execute malicious instructions almost instantaneously, their internal states exhibit abnormally high decision entropy. Motivated by this latent hesitation, we investigate Representation Engineering (RepE) as a robust detection strategy. By extracting hidden states at the tool-input position, we revealed that the RepE-based circuit breaker successfully identifies and intercepts unauthorized actions before the agent commits to them, achieving high detection accuracy across diverse LLM backbones. This study exposes the limitations of current IPI defenses and provides a highly practical paradigm for building resilient multi-agent architectures. --- ### Where to Steer: Input-Dependent Layer Selection for Steering Improves LLM Alignment Authors: Soham Gadgil, Chris Lin, Su-In Lee Published: 2026-04-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.03867v1 Steering vectors have emerged as a lightweight and effective approach for aligning large language models (LLMs) at inference time, enabling modulation over model behaviors by shifting LLM representations towards a target behavior. However, existing methods typically apply steering vectors at a globally fixed layer, implicitly assuming that the optimal intervention layer is invariant across inputs. We argue that this assumption is fundamentally limited, as representations relevant to a target behavior can be encoded at different layers depending on the input. Theoretically, we show that different inputs can require steering at different layers to achieve alignment with a desirable model behavior. We also provide empirical evidence that the optimal steering layer varies substantially across inputs in practice. Motivated by these observations, we introduce Where to Steer (W2S), a framework that adaptively selects the intervention layer conditioned on the input, by learning a mapping from input embeddings to optimal steering layers. Across multiple LLMs and alignment behaviors, W2S consistently outperforms fixed-layer baselines, with improvements in both in-distribution and out-of-distribution settings. Our findings highlight the importance of input-dependent control in LLM alignment and demonstrate that adaptive layer selection is a key design dimension missing in the current methodology of steering vectors. --- ### SecureAFL: Secure Asynchronous Federated Learning Authors: Anjun Gao, Feng Wang, Zhenglin Wan, Yueyang Quan, Zhuqing Liu, Minghong Fang Published: 2026-04-04 Categories: cs.CR, cs.DC, cs.LG arXiv: https://arxiv.org/abs/2604.03862v1 Federated learning (FL) enables multiple clients to collaboratively train a global machine learning model via a server without sharing their private training data. In traditional FL, the system follows a synchronous approach, where the server waits for model updates from numerous clients before aggregating them to update the global model. However, synchronous FL is hindered by the straggler problem. To address this, the asynchronous FL architecture allows the server to update the global model immediately upon receiving any client's local model update. Despite its advantages, the decentralized nature of asynchronous FL makes it vulnerable to poisoning attacks. Several defenses tailored for asynchronous FL have been proposed, but these mechanisms remain susceptible to advanced attacks or rely on unrealistic server assumptions. In this paper, we introduce SecureAFL, an innovative framework designed to secure asynchronous FL against poisoning attacks. SecureAFL improves the robustness of asynchronous FL by detecting and discarding anomalous updates while estimating the contributions of missing clients. Additionally, it utilizes Byzantine-robust aggregation techniques, such as coordinate-wise median, to integrate the received and estimated updates. Extensive experiments on various real-world datasets demonstrate the effectiveness of SecureAFL. --- ### A Bayesian Information-Theoretic Approach to Data Attribution Authors: Dharmesh Tailor, Nicolò Felicioni, Kamil Ciosek Published: 2026-04-04 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2604.03858v2 Training Data Attribution (TDA) seeks to trace model predictions back to influential training examples, enhancing interpretability and safety. We formulate TDA as a Bayesian information-theoretic problem: subsets are scored by the information loss they induce - the entropy increase at a query when removed. This criterion credits examples for resolving predictive uncertainty rather than label noise. To scale to modern networks, we approximate information loss using a Gaussian Process surrogate built from tangent features. We show this aligns with classical influence scores for single-example attribution while promoting diversity for subsets. For even larger-scale retrieval, we relax to an information-gain objective and add a variance correction for scalable attribution in vector databases. Experiments show competitive performance on counterfactual sensitivity, ground-truth retrieval and coreset selection, showing that our method scales to modern architectures while bridging principled measures with practice. --- ### Understanding When Poisson Log-Normal Models Outperform Penalized Poisson Regression for Microbiome Count Data Authors: Daniel Agyapong, Julien Chiquet, Jane Marks, Toby Dylan Hocking Published: 2026-04-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.03853v1 Multivariate count models are often justified by their ability to capture latent dependence, but researchers receive little guidance on when this added structure improves on simpler penalized marginal Poisson regression. We study this question using real microbiome data under a unified held-out evaluation framework. For count prediction, we compare PLN and GLMNet(Poisson) on 20 datasets spanning 32 to 18,270 samples and 24 to 257 taxa, using held-out Poisson deviance under leave-one-taxon-out prediction with 3-fold sample cross-validation rather than synthetic or in-sample criteria. For network inference, we compare PLNNetwork and GLMNet(Poisson) neighborhood selection on five publicly available datasets with experimentally validated microbial interaction truth. PLN outperforms GLMNet(Poisson) on most count-prediction datasets, with gains up to 38 percent. The primary predictor of the winner is the sample-to-taxon ratio, with mean absolute correlation as the strongest secondary signal and overdispersion as an additional predictor. PLNNetwork performs best on broad undirected interaction benchmarks, whereas GLMNet(Poisson) is better aligned with local or directional effects. Taken together, these results provide guidance for choosing between latent multivariate count models and penalized Poisson regression in biological count prediction and interaction recovery. --- ### Collapse-Free Prototype Readout Layer for Transformer Encoders Authors: Giansalvo Cirrincione, Rahul Ranjeev Kumar Published: 2026-04-04 Categories: cs.LG, cs.NE arXiv: https://arxiv.org/abs/2604.03850v1 DDCL-Attention is a prototype-based readout layer for transformer encoders that replaces simple pooling methods, such as mean pooling or class tokens, with a learned compression mechanism. It uses a small set of global prototype vectors and assigns tokens to them through soft probabilistic matching, producing compact token summaries at linear complexity in sequence length. The method offers three main advantages. First, it avoids prototype collapse through an exact decomposition of the training loss into a reconstruction term and a diversity term, ensuring that prototypes remain distinct. Second, its joint training with the encoder is shown to be stable under a practical timescale condition, using Tikhonov's singular perturbation theory and explicit learning-rate constraints. Third, the same framework supports three uses: a final readout layer, a differentiable codebook extending VQ-VAE, and a hierarchical document compressor. Experiments on four datasets confirm the theoretical predictions: the loss decomposition holds exactly, prototype separation grows as expected when the stability condition is met, and the codebook reaches full utilization, outperforming standard hard vector quantization. An additional study on orbital debris classification shows that the method also applies beyond standard NLP and vision tasks, including scientific tabular data. --- ### Explainability-Guided Adversarial Attacks on Transformer-Based Malware Detectors Using Control Flow Graphs Authors: Andrew Wheeler, Kshitiz Aryal, Maanak Gupta Published: 2026-04-04 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2604.03843v1 Transformer-based malware detection systems operating on graph modalities such as control flow graphs (CFGs) achieve strong performance by modeling structural relationships in program behavior. However, their robustness to adversarial evasion attacks remains underexplored. This paper examines the vulnerability of a RoBERTa-based malware detector that linearizes CFGs into sequences of function calls, a design choice that enables transformer modeling but may introduce token-level sensitivities and ordering artifacts exploitable by adversaries. By evaluating evasion strategies within this graph-to-sequence framework, we provide insight into the practical robustness of transformer-based malware detectors beyond aggregate detection accuracy. This paper proposes a white-box adversarial evasion attack that leverages explainability mechanisms to identify and perturb most influential graph components. Using token- and word-level attributions derived from integrated gradients, the attack iteratively replaces positively attributed function calls with synthetic external imports, producing adversarial CFG representations without altering overall program structure. Experimental evaluation on small- and large-scale Windows Portable Executable (PE) datasets demonstrates that the proposed method can reliably induce misclassification, even against models trained to high accuracy. Our results highlight that explainability tools, while valuable for interpretability, can also expose critical attack surfaces in transformer-based malware detectors. --- ### Training a Student Expert via Semi-Supervised Foundation Model Distillation Authors: Pardis Taghavi, Tian Liu, Renjie Li, Reza Langari, Zhengzhong Tu Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03841v1 Foundation models deliver strong perception but are often too computationally heavy to deploy, and adapting them typically requires costly annotations. We introduce a semi-supervised knowledge distillation (SSKD) framework that compresses pre-trained vision foundation models (VFMs) into compact experts using limited labeled and abundant unlabeled data, and instantiate it for instance segmentation where per-pixel labels are particularly expensive. The framework unfolds in three stages: (1) domain adaptation of the VFM(s) via self-training with contrastive calibration, (2) knowledge transfer through a unified multi-objective loss, and (3) student refinement to mitigate residual pseudo-label bias. Central to our approach is an instance-aware pixel-wise contrastive loss that fuses mask and class scores to extract informative negatives and enforce clear inter-instance margins. By maintaining this contrastive signal across both adaptation and distillation, we align teacher and student embeddings and more effectively leverage unlabeled images. On Cityscapes and ADE20K, our $\approx 11\times$ smaller student improves over its zero-shot VFM teacher(s) by +11.9 and +8.6 AP, surpasses adapted teacher(s) by +3.4 and +1.5 AP, and outperforms state-of-the-art SSKD methods on benchmarks. --- ### New insights into Elo algorithm for practitioners and statisticians Authors: Leszek Szczecinski Published: 2026-04-04 Categories: stat.ME, cs.LG arXiv: https://arxiv.org/abs/2604.03840v1 This work reconciles two perspectives on the Elo ranking that coexist in the literature: the practitioner's view as a heuristic feedback rule, and the statistician's view as online maximum likelihood estimation via stochastic gradient ascent. Both perspectives coincide exactly in the binary case (iff the expected score is the logistic function). However, estimation noise forces a principled decoupling between the model used for ranking and the model used for prediction: the effective scale and home-field advantage parameter must be adjusted to account for the noise. We provide both closed-form corrections and a data-driven identification procedure. For multilevel outcomes, an exact relationship exists when outcome scores are uniformly spaced, but approximations are preferred in general: they account for estimation noise and better fit the data. The decoupled approach substantially outperforms the conventional one that reuses the ranking model for prediction, and serves as a diagnostic of convergence status. Applied to six years of FIFA men's ranking, we find that the ranking had not converged for the vast majority of national teams. The paper is written in a semi-tutorial style accessible to practitioners, with all key results accompanied by closed-form expressions and numerical examples. --- ### Beyond Task-Driven Features for Object Detection Authors: Meilun Zhou, Alina Zare Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03839v1 Task-driven features learned by modern object detectors optimize end task loss yet often capture shortcut correlations that fail to reflect underlying annotation structure. Such representations limit transfer, interpretability, and robustness when task definitions change or supervision becomes sparse. This paper introduces an annotation-guided feature augmentation framework that injects embeddings into an object detection backbone. The method constructs dense spatial feature grids from annotation-guided latent spaces and fuses them with feature pyramid representations to influence region proposal and detection heads. Experiments across wildlife and remote sensing datasets evaluate classification, localization, and data efficiency under multiple supervision regimes. Results show consistent improvements in object focus, reduced background sensitivity, and stronger generalization to unseen or weakly supervised tasks. The findings demonstrate that aligning features with annotation geometry yields more meaningful representations than purely task optimized features. --- ### Task-Guided Multi-Annotation Triplet Learning for Remote Sensing Representations Authors: Meilun Zhou, Alina Zare Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03837v1 Prior multi-task triplet loss methods relied on static weights to balance supervision between various types of annotation. However, static weighting requires tuning and does not account for how tasks interact when shaping a shared representation. To address this, the proposed task-guided multi-annotation triplet loss removes this dependency by selecting triplets through a mutual-information criteria that identifies triplets most informative across tasks. This strategy modifies which samples influence the representation rather than adjusting loss magnitudes. Experiments on an aerial wildlife dataset compare the proposed task-guided selection against several triplet loss setups for shaping a representation in an effective multi-task manner. The results show improved classification and regression performance and demonstrate that task-aware triplet selection produces a more effective shared representation for downstream tasks. --- ### Cost-Efficient Multi-Scale Fovea for Semantic-Based Visual Search Attention Authors: João Luzio, Alexandre Bernardino, Plinio Moreno Published: 2026-04-04 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2604.03836v2 Semantics are one of the primary sources of top-down preattentive information. Modern deep object detectors excel at extracting such valuable semantic cues from complex visual scenes. However, the size of the visual input to be processed by these detectors can become a bottleneck, particularly in terms of time costs, affecting an artificial attention system's biological plausibility and real-time deployability. Inspired by classical exponential density roll-off topologies, we apply a new artificial foveation module to our novel attention prediction pipeline: the Semantic-based Bayesian Attention (SemBA) framework. We aim at reducing detection-related computational costs without compromising visual task accuracy, thereby making SemBA more biologically plausible. The proposed multi-scale pyramidal field-of-view retains maximum acuity at an innermost level, around a focal point, while gradually increasing distortion for outer levels to mimic peripheral uncertainty via downsampling. In this work we evaluate the performance of our novel Multi-Scale Fovea, incorporated into SemBA, on target-present visual search. We also compare it against other artificial foveal systems, and conduct ablation studies with different deep object detection models to assess the impact of the new topology in terms of computational costs. We experimentally demonstrate that including the new Multi-Scale Fovea module effectively reduces inherent processing costs while improving SemBA's scanpath prediction accuracy. Remarkably, we show that SemBA closely approximates human consistency while retaining the actual human fovea's proportions. --- ### SPARK-IL: Spectral Retrieval-Augmented RAG for Knowledge-driven Deepfake Detection via Incremental Learning Authors: Hessen Bougueffa Eutamene, Abdellah Zakaria Sellam, Abdelmalik Taleb-Ahmed, Abdenour Hadid Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03833v1 Detecting AI-generated images remains a significant challenge because detectors trained on specific generators often fail to generalize to unseen models; however, while pixel-level artifacts vary across models, frequency-domain signatures exhibit greater consistency, providing a promising foundation for cross-generator detection. To address this, we propose SPARK-IL, a retrieval-augmented framework that combines dual-path spectral analysis with incremental learning by utilizing a partially frozen ViT-L/14 encoder for semantic representations alongside a parallel path for raw RGB pixel embeddings. Both paths undergo multi-band Fourier decomposition into four frequency bands, which are individually processed by Kolmogorov-Arnold Networks (KAN) with mixture-of-experts for band-specific transformations before the resulting spectral embeddings are fused via cross-attention with residual connections. During inference, this fused embedding retrieves the $k$ nearest labeled signatures from a Milvus database using cosine similarity to facilitate predictions via majority voting, while an incremental learning strategy expands the database and employs elastic weight consolidation to preserve previously learned transformations. Evaluated on the UniversalFakeDetect benchmark across 19 generative models -- including GANs, face-swapping, and diffusion methods -- SPARK-IL achieves a 94.6\% mean accuracy, with the code to be publicly released at https://github.com/HessenUPHF/SPARK-IL. --- ### Affording Process Auditability with QualAnalyzer: An Atomistic LLM Analysis Tool for Qualitative Research Authors: Max Hao Lu, Ryan Ellegood, Rony Rodriguez-Ramirez, Sophia Blumert Published: 2026-04-04 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.03820v1 Large language models are increasingly used for qualitative data analysis, but many workflows obscure how analytic conclusions are produced. We present QualAnalyzer, an open-source Chrome extension for Google Workspace that supports atomistic LLM analysis by processing each data segment independently and preserving the prompt, input, and output for every unit. Through two case studies -- holistic essay scoring and deductive thematic coding of interview transcripts -- we show that this approach creates a legible audit trail and helps researchers investigate systematic differences between LLM and human judgments. We argue that process auditability is essential for making LLM-assisted qualitative research more transparent and methodologically robust. --- ### ActivityForensics: A Comprehensive Benchmark for Localizing Manipulated Activity in Videos Authors: Peijun Bao, Anwei Luo, Gang Pan, Alex C. Kot, Xudong Jiang Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03819v1 Temporal forgery localization aims to temporally identify manipulated segments in videos. Most existing benchmarks focus on appearance-level forgeries, such as face swapping and object removal. However, recent advances in video generation have driven the emergence of activity-level forgeries that modify human actions to distort event semantics, resulting in highly deceptive forgeries that critically undermine media authenticity and public trust. To overcome this issue, we introduce ActivityForensics, the first large-scale benchmark for localizing manipulated activity in videos. It contains over 6K forged video segments that are seamlessly blended into the video context, rendering high visual consistency that makes them almost indistinguishable from authentic content to the human eye. We further propose Temporal Artifact Diffuser (TADiff), a simple yet effective baseline that exposes artifact cues through a diffusion-based feature regularizer. Based on ActivityForensics, we introduce comprehensive evaluation protocols covering intra-domain, cross-domain, and open-world settings, and benchmark a wide range of state-of-the-art forgery localizers to facilitate future research. The dataset and code are available at https://activityforensics.github.io. --- ### k-Maximum Inner Product Attention for Graph Transformers and the Expressive Power of GraphGPS Authors: Jonas De Schouwer, Haitz Sáez de Ocáriz Borde, Xiaowen Dong Published: 2026-04-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.03815v2 Graph transformers have shown promise in overcoming limitations of traditional graph neural networks, such as oversquashing and difficulties in modeling long-range dependencies. However, their application to large-scale graphs is hindered by the quadratic memory and computational complexity of the all-to-all attention mechanism. Although alternatives such as linearized attention and restricted attention patterns have been proposed, these often degrade performance or limit expressive power. To better balance efficiency and effectiveness, we introduce k-Maximum Inner Product (k-MIP) attention for graph transformers. k-MIP attention selects the most relevant key nodes per query via a top-k operation, yielding a sparse yet flexible attention pattern. Combined with an attention score computation based on symbolic matrices, this results in linear memory complexity and practical speedups of up to an order of magnitude compared to all-to-all attention, enabling the processing of graphs with over 500k nodes on a single A100 GPU. We provide a theoretical analysis of expressive power, showing that k-MIP attention does not compromise the expressiveness of graph transformers: specifically, we prove that k-MIP transformers can approximate any full-attention transformer to arbitrary precision. In addition, we analyze the expressive power of the GraphGPS framework, in which we integrate our attention mechanism, and establish an upper bound on its graph distinguishing capability in terms of the S-SEG-WL test. Finally, we validate our approach on the Long Range Graph Benchmark, the City-Networks benchmark, and two custom large-scale inductive point cloud datasets, consistently ranking among the top-performing scalable graph transformers. --- ### InCaRPose: In-Cabin Relative Camera Pose Estimation Model and Dataset Authors: Felix Stillger, Lukas Hahn, Frederik Hasecke, Tobias Meisen Published: 2026-04-04 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.03814v1 Camera extrinsic calibration is a fundamental task in computer vision. However, precise relative pose estimation in constrained, highly distorted environments, such as in-cabin automotive monitoring (ICAM), remains challenging. We present InCaRPose, a Transformer-based architecture designed for robust relative pose prediction between image pairs, which can be used for camera extrinsic calibration. By leveraging frozen backbone features such as DINOv3 and a Transformer-based decoder, our model effectively captures the geometric relationship between a reference and a target view. Unlike traditional methods, our approach achieves absolute metric-scale translation within the physically plausible adjustment range of in-cabin camera mounts in a single inference step, which is critical for ICAM, where accurate real-world distances are required for safety-relevant perception. We specifically address the challenges of highly distorted fisheye cameras in automotive interiors by training exclusively on synthetic data. Our model is capable of generalization to real-world cabin environments without relying on the exact same camera intrinsics and additionally achieves competitive performance on the public 7-Scenes dataset. Despite having limited training data, InCaRPose maintains high precision in both rotation and translation, even with a ViT-Small backbone. This enables real-time performance for time-critical inference, such as driver monitoring in supervised autonomous driving. We release our real-world In-Cabin-Pose test dataset consisting of highly distorted vehicle-interior images and our code at https://github.com/felixstillger/InCaRPose. --- ### Representational Collapse in Multi-Agent LLM Committees: Measurement and Diversity-Aware Consensus Authors: Dipkumar Patel Published: 2026-04-04 Categories: cs.LG, cs.AI, cs.MA arXiv: https://arxiv.org/abs/2604.03809v1 Multi-agent LLM committees replicate the same model under different role prompts and aggregate outputs by majority vote, implicitly assuming that agents contribute complementary evidence. We embed each agent's chain-of-thought rationale and measure pairwise similarity: across 100 GSM8K questions with three Qwen2.5-14B agents, mean cosine similarity is 0.888 and effective rank is 2.17 out of 3.0, a failure mode we term representational collapse. DALC, a training-free consensus protocol that computes diversity weights from embedding geometry, reaches 87% on GSM8K versus 84% for self-consistency at 26% lower token cost. Ablation experiments reveal 1-3 point per-protocol run-to-run variance, confirm that hint sharing contributes more than diversity weighting alone, and show that encoder choice strongly modulates collapse severity (cosine 0.908 with mxbai versus 0.888 with nomic) and downstream accuracy. The more robust finding is that collapse is measurable, worsens on harder tasks, and that the choice of embedding proxy is a first-order design decision for any latent communication protocol. --- ### Bridging Restoration and Diagnosis: A Comprehensive Benchmark for Retinal Fundus Enhancement Authors: Xuanzhao Dong, Wenhui Zhu, Xiwen Chen, Hao Wang, Xin Li, Yujian Xiong, Jiajun Cheng, Zhipeng Wang, Shao Tang, Oana Dumitrascu, Yalin Wang Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03806v1 Over the past decade, generative models have demonstrated success in enhancing fundus images. However, the evaluation of these models remains a challenge. A benchmark for fundus image enhancement is needed for three main reasons:(1) Conventional denoising metrics such as PSNR and SSIM fail to capture clinically relevant features, such as lesion preservation and vessel morphology consistency, limiting their applicability in real-world settings; (2) There is a lack of unified evaluation protocols that address both paired and unpaired enhancement methods, particularly those guided by clinical expertise; and (3) An evaluation framework should provide actionable insights to guide future advancements in clinically aligned enhancement models. To address these gaps, we introduce EyeBench-V2, a benchmark designed to bridge the gap between enhancement model performance and clinical utility. Our work offers three key contributions:(1) Multi-dimensional clinical-alignment through downstream evaluations: Beyond standard enhancement metrics, we assess performance across clinically meaningful tasks including vessel segmentation, diabetic retinopathy (DR) grading, generalization to unseen noise patterns, and lesion segmentation. (2) Expert-guided evaluation design: We curate a novel dataset enabling fair comparisons between paired and unpaired enhancement methods, accompanied by a structured manual assessment protocol by medical experts, which evaluates clinically critical aspects such as lesion structure alterations, background color shifts, and the introduction of artificial structures. (3) Actionable insights: Our benchmark provides a rigorous, task-oriented analysis of existing generative models, equipping clinical researchers with the evidence needed to make informed decisions, while also identifying limitations in current methods to inform the design of next-generation enhancement models. --- ### Rényi Attention Entropy for Patch Pruning Authors: Hiroaki Aizawa, Yuki Igaue Published: 2026-04-04 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2604.03803v1 Transformers are strong baselines in both vision and language because self-attention captures long-range dependencies across tokens. However, the cost of self-attention grows quadratically with the number of tokens. Patch pruning mitigates this cost by estimating per-patch importance and removing redundant patches. To identify informative patches for pruning, we introduce a criterion based on the Shannon entropy of the attention distribution. Low-entropy patches, which receive selective and concentrated attention, are kept as important, while high-entropy patches with attention spread across many locations are treated as redundant. We also extend the criterion from Shannon to Rényi entropy, which emphasizes sharp attention peaks and supports pruning strategies that adapt to task needs and computational limits. In experiments on fine-grained image recognition, where patch selection is critical, our method reduced computation while preserving accuracy. Moreover, adjusting the pruning policy through the Rényi entropy measure yields further gains and improves the trade-off between accuracy and computation. --- ### HistoFusionNet: Histogram-Guided Fusion and Frequency-Adaptive Refinement for Nighttime Image Dehazing Authors: Mohammad Heydari, Wei Dong, Shahram Shirani, Jun Chen, Han Zhou Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03800v1 Nighttime image dehazing remains a challenging low-level vision problem due to the joint presence of haze, glow, non-uniform illumination, color distortion, and sensor noise, which often invalidate assumptions commonly used in daytime dehazing. To address these challenges, we propose HistoFusionNet, a transformer-enhanced architecture tailored for nighttime image dehazing by combining histogram-guided representation learning with frequency-adaptive feature refinement. Built upon a multi-scale encoder-decoder backbone, our method introduces histogram transformer blocks that model long-range dependencies by grouping features according to their dynamic-range characteristics, enabling more effective aggregation of similarly degraded regions under complex nighttime lighting. To further improve restoration fidelity, we incorporate a frequency-aware refinement branch that adaptively exploits complementary low- and high-frequency cues, helping recover scene structures, suppress artifacts, and enhance local details. This design yields a unified framework that is particularly well suited to the heterogeneous degradations encountered in real nighttime hazy scenes. Extensive experiments and highly competitive performance of our method on the NTIRE 2026 Nighttime Image Dehazing Challenge benchmark demonstrate the effectiveness of the proposed method. Our team ranked 1st among 22 participating teams, highlighting the robustness and competitive performance of HistoFusionNet. The code is available at: https://github.com/heydarimo/Night-Time-Dehazing --- ### Next-Scale Autoregressive Models for Text-to-Motion Generation Authors: Zhiwei Zheng, Shibo Jin, Lingjie Liu, Mingmin Zhao Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03799v2 Autoregressive (AR) models offer stable and efficient training, but standard next-token prediction is not well aligned with the temporal structure required for text-conditioned motion generation. We introduce MoScale, a next-scale AR framework that generates motion hierarchically from coarse to fine temporal resolutions. By providing global semantics at the coarsest scale and refining them progressively, MoScale establishes a causal hierarchy better suited for long-range motion structure. To improve robustness under limited text-motion data, we further incorporate cross-scale hierarchical refinement for improving per-scale initial predictions and in-scale temporal refinement for selective bidirectional re-prediction. MoScale achieves SOTA text-to-motion performance with high training efficiency, scales effectively with model size, and generalizes zero-shot to diverse motion generation and editing tasks. --- ### Confidence-Driven Facade Refinement of 3D Building Models Using MLS Point Clouds Authors: Xiaoyu Huang Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03797v1 Digital twins require continuous maintenance to meet the increasing demand for high-precision geospatial data. However, traditional coarse CityGML building models, typically derived from Airborne Laser Scanning (ALS), often exhibit significant geometric deficiencies, particularly regarding facade accuracy due to the nadir perspective of airborne sensors. Integrating these coarse models with high-precision Mobile Laser Scanning (MLS) data is essential to recover detailed facade geometry. Unlike reconstruction-from-scratch approaches that discard existing semantic information and rely heavily on complete data coverage, this work presents an automated refinement framework that utilizes the coarse model as a geometric prior. This method enables targeted updates to facade geometry even in complex urban environments. It integrates surface matching to identify outdated surfaces and employs a binary integer optimization to select optimal faces from candidate data. Crucially, hard constraints are enforced within the optimization to ensure the topological validity of the refined output. Experimental results demonstrate that the proposed approach effectively corrects facade misalignments, reducing the Cloud-to-Mesh RMSE by approximately 36% and achieving centimeter-level alignment. Furthermore, the framework guarantees strictly watertight and manifold geometry, providing a robust solution for upgrading ALS-derived city models. --- ### Automated Conjecture Resolution with Formal Verification Authors: Haocheng Ju, Guoxiong Gao, Jiedong Jiang, Bin Wu, Zeming Sun, Shurui Liu, Leheng Chen, Yutong Wang, Yuefeng Wang, Zichen Wang, Wanyi He, Peihao Wu et al. Published: 2026-04-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.03789v2 Recent advances in large language models have significantly improved their ability to perform mathematical reasoning, extending from elementary problem solving to increasingly capable performance on research-level problems. However, reliably solving and verifying such problems remains challenging due to the inherent ambiguity of natural language reasoning. In this paper, we propose an automated framework that integrates natural language reasoning with formal verification to tackle research-level mathematical problems. Our framework consists of two components: an informal reasoning agent, Rethlas, and a formal verification agent, Archon. Rethlas combines reasoning primitives with our theorem search engine, Matlas, to explore solution strategies and construct candidate proofs. Archon, equipped with LeanSearch, translates informal arguments into formalized Lean 4 projects through task decomposition, iterative refinement, and automated proof synthesis, ensuring machine-checkable correctness. Using this framework, we resolve an open problem in commutative algebra and formally verify the resulting proof in Lean 4 with essentially no human involvement. Additional case studies illustrate the capabilities of Rethlas in informal mathematical reasoning and discovery, as well as the ability of Archon to formalize research-level proofs in Lean 4. Our experiments demonstrate that strong theorem retrieval tools enable the discovery and application of cross-domain mathematical techniques, while the formal agent can autonomously fill nontrivial gaps in informal arguments. More broadly, our work illustrates a promising paradigm for mathematical research in which informal and formal reasoning systems, equipped with theorem retrieval tools, operate in tandem to produce verifiable results, reduce human effort, and support human-AI collaborative mathematical research. --- ### On the Efficiency of Sinkhorn-Knopp for Entropically Regularized Optimal Transport Authors: Kun He Published: 2026-04-04 Categories: cs.DS, cs.LG arXiv: https://arxiv.org/abs/2604.03787v1 The Sinkhorn--Knopp (SK) algorithm is a cornerstone method for matrix scaling and entropically regularized optimal transport (EOT). Despite its empirical efficiency, existing theoretical guarantees to achieve a target marginal accuracy $\varepsilon$ deteriorate severely in the presence of outliers, bottlenecked either by the global maximum regularized cost $η\|C\|_\infty$ (where $η$ is the regularization parameter and $C$ the cost matrix) or the matrix's minimum-to-maximum entry ratio $ν$. This creates a fundamental disconnect between theory and practice. In this paper, we resolve this discrepancy. For EOT, we introduce the novel concept of well-boundedness, a local bulk mass property that rigorously isolates the well-behaved portion of the data from extreme outliers. We prove that governed by this fundamental notion, SK recovers the target transport plan for a problem of dimension $n$ in $O(\log n - \log \varepsilon)$ iterations, completely independent of the regularized cost $η\|C\|_\infty$. Furthermore, we show that a virtually cost-free pre-scaling step eliminates the dimensional dependence entirely, accelerating convergence to a strictly dimension-free $O(\log(1/\varepsilon))$ iterations. Beyond EOT, we establish a sharp phase transition for general $(\boldsymbol{u},\boldsymbol{v})$-scaling governed by a critical matrix density threshold. We prove that when a matrix's density exceeds this threshold, the iteration complexity is strictly independent of $ν$. Conversely, when the density falls below this threshold, the dependence on $ν$ becomes unavoidable; in this sub-critical regime, we construct instances where SK requires $Ω(n/\varepsilon)$ iterations. --- ### Communication Gain and Delay Cost Under Cross-Timestep Delays in Cooperative Multi-Agent Reinforcement Learning Authors: Zihong Gao, Hongjian Liang, Lei Hao, Liangjun Ke Published: 2026-04-04 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2604.03785v3 Communication is essential for coordination in \emph{cooperative} multi-agent reinforcement learning under partial observability, yet \emph{cross-timestep} delays cause messages to arrive multiple timesteps after generation, inducing temporal misalignment and making information stale when consumed. We formalize this setting as a delayed-communication partially observable Markov game (DeComm-POMG) and decompose a message's effect into \emph{communication gain} and \emph{delay cost}, yielding the Communication Gain and Delay Cost (CGDC) metric. We further establish a value-loss bound showing that the degradation induced by delayed messages is upper-bounded by a discounted accumulation of an information gap between the action distributions induced by timely versus delayed messages. Guided by CGDC, we propose \textbf{CDCMA}, an actor--critic framework that requests messages only when predicted CGDC is positive, predicts future observations to reduce misalignment at consumption, and fuses delayed messages via CGDC-guided attention. Experiments on no-teammate-vision variants of Cooperative Navigation and Predator Prey, and on SMAC maps across multiple delay levels show consistent improvements in performance, robustness, and generalization, with ablations validating each component. --- ### Communication Gain and Delay Cost Under Cross-Timestep Delays in Cooperative Multi-Agent Reinforcement Learning Authors: Zihong Gao, Hongjian Liang, Lei Hao, Liangjun Ke Published: 2026-04-04 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2604.03785v2 Communication is essential for coordination in \emph{cooperative} multi-agent reinforcement learning under partial observability, yet \emph{cross-timestep} delays cause messages to arrive multiple timesteps after generation, inducing temporal misalignment and making information stale when consumed. We formalize this setting as a delayed-communication partially observable Markov game (DeComm-POMG) and decompose a message's effect into \emph{communication gain} and \emph{delay cost}, yielding the Communication Gain and Delay Cost (CGDC) metric. We further establish a value-loss bound showing that the degradation induced by delayed messages is upper-bounded by a discounted accumulation of an information gap between the action distributions induced by timely versus delayed messages. Guided by CGDC, we propose \textbf{CDCMA}, an actor--critic framework that requests messages only when predicted CGDC is positive, predicts future observations to reduce misalignment at consumption, and fuses delayed messages via CGDC-guided attention. Experiments on no-teammate-vision variants of Cooperative Navigation and Predator Prey, and on SMAC maps across multiple delay levels show consistent improvements in performance, robustness, and generalization, with ablations validating each component. --- ### An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent Authors: Anja Surina, Arun Suggala, George Tsoukalas, Anton Kovsharov, Sergey Shirobokov, Francisco J. R. Ruiz, Pushmeet Kohli, Swarat Chaudhuri Published: 2026-04-04 Categories: math.OC, cs.AI arXiv: https://arxiv.org/abs/2604.03782v1 We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved exact $\mathcal{O}(1/t)$ rate is achievable. In this work, we resolve this question in the affirmative. This result was discovered autonomously by an AI system capable of writing formal proofs in Lean. The Lean proof can be accessed at https://github.com/google-deepmind/formal-conjectures/pull/3675/commits/a13226b49fd3b897f4c409194f3bcbeb96a08515 --- ### CountsDiff: A Diffusion Model on the Natural Numbers for Generation and Imputation of Count-Based Data Authors: Renzo G. Soatto, Anders Hoel, Greycen Ren, Shorna Alam, Stephen Bates, Nikolaos P. Daskalakis, Caroline Uhler, Maria Skoularidou Published: 2026-04-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.03779v2 Diffusion models have excelled at generative tasks for both continuous and token-based domains, but their application to discrete ordinal data remains underdeveloped. We present CountsDiff, a diffusion framework designed to model distributions on the natural numbers. CountsDiff extends the Blackout diffusion framework by simplifying its formulation through a direct parameterization in terms of a survival probability schedule and an explicit loss weighting. This introduces flexibility through design parameters with direct analogues in existing diffusion modeling frameworks. Beyond this reparameterization, CountsDiff introduces features from modern diffusion models, previously absent in counts-based domains, including continuous-time training, classifier-free guidance, and churn/remasking reverse dynamics that allow non-monotone reverse trajectories. We propose an initial instantiation of CountsDiff and validate it on natural image datasets (CIFAR-10, CelebA), exploring the effects of the introduced design parameters in a complex, well-studied, and interpretable data domain. We then highlight biological count assays as a natural use case, evaluating CountsDiff on single-cell RNA-seq imputation in fetal and heart cell atlases. Remarkably, we find that even this simple instantiation matches or surpasses the performance of a state-of-the-art discrete generative model and leading scRNA-seq imputation methods, while leaving substantial headroom for further gains through optimized design choices in future work. --- ### Cross-Spectral Witness for Hidden Nonequilibrium Beyond the Scalar Ceiling Authors: Yuda Bi, Vince D Calhoun Published: 2026-04-04 Categories: cond-mat.stat-mech, stat.ML arXiv: https://arxiv.org/abs/2604.03775v3 Partial observation is a pervasive obstacle in nonequilibrium physics: coarse graining may absorb hidden forcing into an apparently equilibrium-like reduced description, so a driven system can look reversible through the only variables one can measure. For scalar Gaussian observables of linear stochastic systems, no time-irreversibility statistic can detect the underlying drive. The Lucente--Crisanti ceiling constrains what one channel carries; what two channels carry is a different question, with a sharp closed-form answer. Two simultaneously observed channels retain an off-diagonal cross-spectral sector inaccessible to any scalar reduction; under channel-separable multiplicative structure the observed-channel response factors cancel identically, leaving a closed-form cross-spectral witness controlled only by the hidden spectrum, the loadings, and the innovation scales, strictly positive at every nonzero cross-coupling including at exact timescale coalescence where every scalar reduction is blind. Within general CSM this certifies shared hidden-sector drive; under the additional one-way coupling assumption the witness identifies the total entropy production rate at leading order with a square-root scaling. --- ### When Does Multimodal AI Help? Diagnostic Complementarity of Vision-Language Models and CNNs for Spectrum Management in Satellite-Terrestrial Networks Authors: Yuanhang Li Published: 2026-04-04 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.03774v1 The adoption of vision-language models (VLMs) for wireless network management is accelerating, yet no systematic understanding exists of where these large foundation models outperform lightweight convolutional neural networks (CNNs) for spectrum-related tasks. This paper presents the first diagnostic comparison of VLMs and CNNs for spectrum heatmap understanding in non-terrestrial network and terrestrial network (NTN-TN) cooperative systems. We introduce SpectrumQA, a benchmark comprising 108K visual question-answer pairs across four granularity levels: scene classification (L1), regional reasoning (L2), spatial localization (L3), and semantic reasoning (L4). Our experiments on three NTN-TN scenarios with a frozen Qwen2-VL-7B and a trained ResNet-18 reveal a clear taskdependent complementarity: CNN achieves 72.9% accuracy at severity classification (L1) and 0.552 IoU at spatial localization (L3), while VLM uniquely enables semantic reasoning (L4) with F1=0.576 using only three in-context examples-a capability fundamentally absent in CNN architectures. Chain-of-thought (CoT) prompting further improves VLM reasoning by 12.6% (F1: 0.209->0.233) while having zero effect on spatial tasks, confirming that the complementarity is rooted in architectural differences rather than prompting limitations. A deterministic task-type router that delegates supervised tasks to CNN and reasoning tasks to VLM achieves a composite score of 0.616, a 39.1% improvement over CNN alone. We further show that VLM representations exhibit stronger cross-scenario robustness, with smaller performance degradation in 5 out of 6 transfer directions. These findings provide actionable guidelines: deploy CNNs for spatial localization and VLMs for semantic spectrum reasoning, rather than treating them as substitutes. --- ### M2StyleGS: Multi-Modality 3D Style Transfer with Gaussian Splatting Authors: Xingyu Miao, Xueqi Qiu, Haoran Duan, Yawen Huang, Xian Wu, Jingjing Deng, Yang Long Published: 2026-04-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03773v1 Conventional 3D style transfer methods rely on a fixed reference image to apply artistic patterns to 3D scenes. However, in practical applications such as virtual or augmented reality, users often prefer more flexible inputs, including textual descriptions and diverse imagery. In this work, we introduce a novel real-time styling technique M2StyleGS to generate a sequence of precisely color-mapped views. It utilizes 3D Gaussian Splatting (3DGS) as a 3D presentation and multi-modality knowledge refined by CLIP as a reference style. M2StyleGS resolves the abnormal transformation issue by employing a precise feature alignment, namely subdivisive flow, it strengthens the projection of the mapped CLIP text-visual combination feature to the VGG style feature. In addition, we introduce observation loss, which assists in the stylized scene better matching the reference style during the generation, and suppression loss, which suppresses the offset of reference color information throughout the decoding process. By integrating these approaches, M2StyleGS can employ text or images as references to generate a set of style-enhanced novel views. Our experiments show that M2StyleGS achieves better visual quality and surpasses the previous work by up to 32.92% in terms of consistency. --- ### Debiased Machine Learning for Conformal Prediction of Counterfactual Outcomes Under Runtime Confounding Authors: Keith Barnatchez, Kevin P. Josey, Rachel C. Nethery, Giovanni Parmigiani Published: 2026-04-04 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2604.03772v1 Data-driven decision making frequently relies on predicting counterfactual outcomes. In practice, researchers commonly train counterfactual prediction models on a source dataset to inform decisions on a possibly separate target population. Conformal prediction has arisen as a popular method for producing assumption-lean prediction intervals for counterfactual outcomes that would arise under different treatment decisions in the target population of interest. However, existing methods require that every confounding factor of the treatment-outcome relationship used for training on the source data is additionally measured in the target population, risking miscoverage if important confounders are unmeasured in the target population. In this paper, we introduce a computationally efficient debiased machine learning framework that allows for valid prediction intervals when only a subset of confounders is measured in the target population, a common challenge referred to as runtime confounding. Grounded in semiparametric efficiency theory, we show the resulting prediction intervals achieve desired coverage rates with faster convergence compared to standard methods. Through numerous synthetic and semi-synthetic experiments, we demonstrate the utility of our proposed method. ---