arXiv:2605.17363v1 Announce Type: cross
Abstract: Single-layer molybdenum disulfide ($MoS_2$) possesses significant potential for nanoscale optoelectronics, but achieving high-intensity, long-term-stable photoluminescence (PL) emission remains a challenge. In this work, we demonstrate a remarkably robust, more than 8-fold maximum enhancement in the PL intensity of exfoliated and CVD-grown single-layer $MoS_2$ via a non-destructive ultraviolet (UV) laser treatment method. This substantial increase in radiative efficiency is accompanied by a trion-to-neutral exciton transition in the PL signal and a corresponding blue shift of the Raman $E_{2g}^1$ and $A_{1g}$ vibrational modes, signaling successful electron depletion (p-doping) and formation of Mo-O bonds, respectively. Furthermore, we demonstrate precise spatial control over PL properties by confining PL treatment exclusively to the UV laser-treated area. Crucially, the enhanced PL performance shows exceptional longevity; the CVD sample and the exfoliated sample remained stable for the entire monitoring period (72 and 32 days, respectively) under ambient conditions. We further investigated UV laser treatment in a controlled-environment chamber under argon, nitrogen, and oxygen atmospheres, distinguishing the influence of oxygen as the PL treatment agent. These findings establish a reliable pathway for the permanent treatment of single-layer $MoS_2$ PL properties, an essential step toward practical, high-performance nanophotonic devices.
Science Journals
arXiv:2605.17621v1 Announce Type: cross
Abstract: Stochastic forces are usually postulated or obtained by eliminating environmental degrees of freedom. Here we identify a variational origin: fluctuating endpoint data in Hamilton's principle induce fluctuations of the on-shell action. Hamilton--Jacobi propagation transports this boundary imprint, whose gradient generates an effective Langevin force inherited from boundary-action fluctuations. The resulting force is not freely specifiable: its amplitude is filtered by the Hessian of Hamilton's principal function, yielding multiplicative and state-dependent noise. Homogeneous additive Langevin forcing is recovered only as a Markovian coarse-grained limit.
arXiv:2605.15641v2 Announce Type: replace
Abstract: Large language models (LLMs) are increasingly used as general planners in embodied intelligence, enabling high level coordination and low level task planning for both single robot and multi-robot collaboration. This increasing reliance on embodied LLM planners also raises critical security concerns, since misaligned or manipulated instructions can be translated into physical actions. Prior work has studied such threats in single robot settings, while security risks in LLM controlled multi-robot collaboration, especially those propagated through inter robot communication, remain largely unexplored. To bridge this gap, we propose a novel attack paradigm for multi-robot system in which the adversary interacts with only a single entry robot. The compromised robot then propagates malicious intent through peer communication, leading to coordinated unsafe actions across the system. Our evaluation, covering high risk dimensions of dereliction of duty, privacy compromise, and public safety hazards, reveals a persistent safety alignment gap in multi-robot planners. We quantify this process with three metrics, obedience, infectiousness, and stealthiness. Experiments demonstrate both persistent attacker control and rapid propagation: obedience reaches 1.00 in the strongest cases, and infectiousness rises to 0.90. Notably, the attack is highly efficient, requiring as few as 3.0 rounds to compromise all the robots while maintaining a stealthiness score of 0.81. Such risks are amplified when robots must resolve trade offs in critical situations, such as emergencies or conflicts of rights, because the coordination mechanism can unintentionally allow adversarial instructions to override safety requirements. The code is available at https://github.com/TheFatInsect/InfectBot.
arXiv:2605.18120v1 Announce Type: cross
Abstract: Future 6G systems are expected to exploit upper midband spectrum in frequency range 3 (FR3) not only for high throughput communications, but also for sensing services such as localization, detection, and situational awareness. The following paper develops a concrete path from today's coverage-oriented deployments to FR3 networks that treat sensing as a native function. We first show how existing FR2 radars can be time-multiplexed and coordinated under a $6$G medium access control as radar-as-a-service, forming a bridge between legacy sensing and network-managed integrated sensing and communications (ISAC). We then propose a hierarchical FR3 beam-alignment strategy in which coarse access occurs at lower frequencies and refinement occurs at upper FR3, and quantify the resulting sensing and communication capabilities via range-angle Cram{\'{e}}r-Rao bounds in the near field. We identify intra- and inter-beam squint phenomena specific to wideband FR3 arrays, and discuss design approaches to mitigate them. On the signal-processing side, we argue that FR3 sensing cannot rely solely on pilot resources and discuss how much sensing information can be extracted from payload resource elements. We further highlight the role of calibrated FR3 channel simulators and real-time models as the core of wireless digital twins for training and evaluating ISAC algorithms, and discuss how massive MIMO and dense or distributed deployments at FR3 naturally act as large reconfigurable sensor arrays.
arXiv:2603.00652v4 Announce Type: replace-cross
Abstract: Semi-classical analysis is used to investigate synchronous quantum tunneling in a multidimensional potential energy surface (PES) characterized by four degenerate minima, serving as a foundational model for coupled vibrational modes. The primary challenge in such systems is the non-linear ``locking" of trajectories where degrees of freedom must traverse their respective barriers synchronously. Starting from the Feynman path integral in imaginary time, we analytically identify longitudinal, transverse, and diagonal instanton configurations that mediate competing tunneling pathways between minima. The translational zero mode for each trajectory is treated rigorously by transforming to a comoving rotating frame. By applying the Gelfand-Yaglom method to the functional determinant and utilizing graph theory to sum the multi-flavor dilute instanton gas , we derive coherent Rabi-type oscillations and exact ground-state tunneling splittings. Crucially, we identify a critical coupling regime where the discrete $D_4$ spatial symmetry of the minima undergoes a topological 'melting' transition into a continuous $O(2)$ rotational symmetry. These analytical results, validated against high-precision numerical diagonalization, provide a rigorous benchmark for multidimensional computational techniques, such as Ring Polymer Instanton (RPI) theory, particularly in the strongly coupled regime where standard discrete instanton approximations break down.
arXiv:2605.18660v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly used as collaborative assistants, yet dominant NLP evaluation practices remain centered on aggregate metrics such as accuracy and fluency. These approaches often overlook behaviors that are critical in human-facing settings (e.g., consistency across multiple turns and iterative refinement). In this paper, we examine limitations of current NLP evaluation practices and introduce TCR, a structured framework for evaluating human--AI interaction using educational LLM assistants as an illustrative example. TCR emphasizes dimensions such as transparency, consistency, and refinement. We further present structured evaluation prompts and illustrative interaction examples demonstrating how structured evaluation can complement aggregate metrics and LLM-as-a-judge approaches. Our work highlights the need for more human-centered evaluation practices for interactive LLM systems.
arXiv:2605.18169v1 Announce Type: cross
Abstract: Multi-controlled Toffoli gates are fundamental building blocks in quantum computation, with applications in quantum arithmetic, simulation, and search algorithms. In fault-tolerant architectures, their realization is constrained by the high cost of non-Clifford resources, particularly in terms of T-count and T-depth. Recent advances have demonstrated that the use of ancillary qubits, relative-phase Toffoli gates, and dynamic circuit techniques can substantially reduce this overhead.
In this work, we investigate the decomposition of large Toffoli gates using 3- and 4-input relative-phase Toffoli gates in the presence of a single clean ancilla and conditionally clean ancillas. We derive explicit resource bounds for Clifford+T implementations incorporating dynamic-circuit-based uncomputation and measurement-conditioned corrections. Our analysis emphasizes T-depth reduction under fixed CX and T-count overhead, ensuring relevance for near-term devices. We show that introducing 4-input relative-phase Toffoli gates enables significant T-depth reductions through enhanced parallelism while maintaining favorable ancilla requirements. We further validate our theoretical results through experimental evaluation and comparative analysis with existing approaches.
arXiv:2305.07152v4 Announce Type: replace
Abstract: Robotic assisted (RA) surgery promises to transform surgical intervention. Intuitive Surgical is committed to fostering these changes and the machine learning models and algorithms that will enable them. With these goals in mind we have invited the surgical data science community to participate in a yearly competition hosted through the Medical Imaging Computing and Computer Assisted Interventions (MICCAI) conference. With varying changes from year to year, we have challenged the community to solve difficult machine learning problems in the context of advanced RA applications. Here we document the results of these challenges, focusing on surgical tool localization (SurgToolLoc) and surgical visual understanding (SurgVU). The publicly released dataset that accompanies these challenges is detailed in a separate paper arXiv:2501.09209 [1].
arXiv:2605.18662v1 Announce Type: new
Abstract: Noise-tolerant PAC learning of linear models has been of central interests in machine learning community since the last century. In recent years, many computationally-efficient algorithms have been proposed for the problem of learning linear threshold functions under multiple noise models. Yet, when the problem is considered under multiclass learning settings, i.e. when the number of classes $k$ is at least $3$, it is unknown whether there exist computationally-efficient PAC learning algorithms when the data sets are maliciously corrupted. In this paper, we consider that the marginal distribution is a mixture of bounded variance distributions and the data sets satisfy a margin condition at the same time. We show that there exists a computationally-efficient algorithm that PAC learns multiclass linear classifiers $\{h_w:x\mapsto \arg\max_{y\in[k]}w_y\cdot x, x\in \mathbb{R}^d, w\in\mathbb{R}^{kd}\}$ using at most $O(k^2\cdot (d\log d+\log k))$ samples even under a constant rate of nasty noise. Our algorithm consists of two main ingredients: a cluster-based pruning scheme and a standard multiclass hinge loss minimization program. Even in the special case of binary setting, i.e. $k=2$, our result is strictly stronger than all prior works.
arXiv:2309.11965v4 Announce Type: replace
Abstract: In this paper, coordination control of discrete event systems under joint sensor and actuator attacks is investigated. Sensor attacks are described by a set of attack languages using a proposed ALTER model. Several local supervisors are used to control the system. The goal is to design local supervisors to ensure safety of the system even under cyber attacks (CA). The necessary and sufficient conditions for the existence of such supervisors are derived in terms of conditional decomposability, CA-controllability and CA-observability. A method is developed to calculate local state estimates under sensor attacks. Two methods are also developed to design local supervisors, one for discrete event systems satisfying conditional decomposability, CA-controllability and CA-observability, and one for discrete event systems satisfying conditional decomposability only. The approach works for both stealthy and non-stealthy attacks. A practical example is given to illustrate the results.
arXiv:2605.18664v1 Announce Type: new
Abstract: Atom-based quantum computing exploits the ability to enhance atom-atom interactions by employing laser excitation to higher-excited Rydberg states. Additional fields that drive transitions between Rydberg states can offer independent control of these atom-atom interactions. However, as microwave (mw) fields only provide access to states with similar principal quantum number $n$, their ability to switch the interactions' strength is limited. Here, we use a pulsed terahertz field to rapidly switch the strength of interactions between Rydberg atoms by three orders of magnitude. We demonstrate interaction switching using photon storage, where the terahertz field induces an interaction induced dephasing of the stored photon. This ability to switch interactions offers advantages for single-qubit readout, state-detection schemes, quantum annealing, and Rydberg quantum optics.
arXiv:2605.18667v1 Announce Type: new
Abstract: Earth embedding models transform Earth observation data into embeddings uniquely tied to locations on the Earth's surface. These models are typically evaluated in isolation, comparing the downstream task performance across different Earth embeddings. However, spatially aligned embeddings can naturally be fused, providing richer information per location, a capability that isolated evaluations fail to capture. We therefore propose assessing Earth embeddings by their complementarity: the performance gain of fused embeddings over the best single-model baseline. To operationalise this, we introduce an embedding complementarity index applicable to any embedding and task, and evaluate four Earth embedding models (AlphaEarth, Tessera, GeoCLIP, SatCLIP) in isolation, in all pairs, and jointly across six downstream tasks. Fused embeddings outperform the best single model in four out of six tasks, confirming that single-embedding evaluations often underestimate Earth embedding capabilities. Complementarity proves both task- and location-dependent. Further, for a land cover regression task, we find that complementarity is partially determined by the spatial scale of land cover classes. Complementarity reframes Earth embeddings: the greatest future gains may come not from any single Earth embedding model, but from combinations that are better together.
arXiv:2503.14800v3 Announce Type: replace
Abstract: Effective long-term memory management is crucial for language models handling extended contexts. We introduce the Enhanced Ranked Memory Augmented Retrieval (ERMAR) framework, which dynamically ranks memory entries based on relevance. Unlike prior models, ERMAR employs a novel relevance scoring mechanism and a pointwise re-ranking model for key-value embeddings, inspired by learning-to-rank techniques in information retrieval. By integrating historical usage patterns and adaptive retrieval, ERMAR achieves state-of-the-art results on standard benchmarks, demonstrating superior scalability and performance in long-context tasks.
arXiv:2505.20218v2 Announce Type: replace
Abstract: Accurate and safe medication recommendations are critical for effective clinical decision-making, especially in multimorbidity cases. However, existing systems rely on point-wise prediction paradigms that overlook synergistic drug effects and potential adverse drug-drug interactions (DDIs). We propose FLAME, a fine-grained list-wise alignment framework for large language models (LLMs), enabling drug-by-drug generation of drug lists. FLAME formulates recommendation as a sequential decision process, where each step adds or removes a single drug. To provide fine-grained learning signals, we devise step-wise Group Relative Policy Optimization (GRPO) with potential-based reward shaping, which explicitly models DDIs and optimizes the contribution of each drug to the overall prescription. Furthermore, FLAME enhances patient modeling by integrating structured clinical knowledge and collaborative information into the representation space of LLMs. Experiments on benchmark datasets demonstrate that FLAME achieves state-of-the-art performance, delivering superior accuracy, controllable safety-accuracy trade-offs, and strong generalization across diverse clinical scenarios. Our code is available at https://github.com/cxfann/Flame.
arXiv:2605.18668v1 Announce Type: new
Abstract: Interactions of relativistically intense laser pulses with structured targets have long been explored for controlling energy absorption and particle acceleration. However, at upcoming multi-petawatt laser facilities, the survivability of such nanostructures under realistic temporal contrast conditions remains a key concern. We report an experimental and simulation study of nanowire targets irradiated by the ELI-NP 1-PW laser without a plasma mirror. At the built in, readily available contrast of $10^{-10}$, the nanowires survive the laser pre-pulse and produce a robust enhancement in relativistic electron flux, energy, and directional emission compared to flat targets indicating that at better contrasts they can show similar enhancement at the 10 PW level. These results establish nanowire targets as resilient and reliable tools for relativistic electron manipulation at state of the art facilities.
arXiv:2605.18670v1 Announce Type: new
Abstract: Risk-limiting audits (RLAs) are post-election auditing procedures that rigorously guarantee a specified maximum probability that an incorrect electoral outcome will not be detected. Aside from ready access to physical ballots, known RLAs require a software-independent accounting of the sizes of each ballot batch, called a ballot manifest. While typical electoral procedures automatically provide rough estimates for batch sizes, even slight inaccuracies (commensurate with the margin of the contest under audit) completely invalidate conventional RLAs (Lindeman et al., EVT 2012). Thus, establishing a sufficiently accurate manifest often requires handling every ballot and can be the dominant cost of conducting the RLA.
We propose two new risk-limiting techniques: 1) A statistical mechanism for ensuring that the batch sizes reported by an untrusted tabulation are, in fact, an accurate manifest; this effectively bootstraps from a rough manifest to an accurate one with sublinear effort. 2) We propose a new class of RLAs called direct ballot selection. This method reverses the traditional comparison procedure and compares uniformly selected ballots against their cast vote records, requiring a new statistical test for identifier duplication but efficiently supporting elections without in order identifiers.
These techniques reduce the complexity of RLAs across many elections. Our two main findings are as follows: 1) The time to create a manifest can be drastically reduced with a modest increase in the number of ballots sampled in the audit. At a 3% margin and a large population, there is a reduction in the overall audit time of at least an order of magnitude across methods. 2) Direct ballot selection improves over state-of-the-art polling for small margins. For Connecticut (29th in population) at a 1% margin, it beats Minerva (Security 2022) by 55% in ballot sample complexity.
arXiv:2508.07292v3 Announce Type: replace
Abstract: Endoscopic diagnosis is an iterative process in which clinicians progressively acquire, compare, and verify local visual evidence before reaching a conclusion. Current AI systems do not adequately support this process because fine-grained evidence acquisition and multi-step reasoning remain weakly coupled. This gives rise to two failure modes, hallucinated evidence and uncorrected error accumulation, that undermine diagnostic reliability. We propose EndoCogniAgent, a closed-loop agentic framework that formulates endoscopic diagnosis as a controlled state update process. At each reasoning round, a central planner selects the next evidence acquisition action, specialized expert tools extract the corresponding observation, and a self-consistency validation mechanism examines the observation along two dimensions, knowledge consistency against the input image and temporal consistency with prior validated findings, before updating the diagnostic state. Validated observations are admitted into the evolving state to condition subsequent planning, while insufficiently supported findings are retained with corrective feedback that redirects the planner toward additional verification. We further introduce EndoAgentBench, a workflow-oriented benchmark comprising 6,132 question-answer pairs from 11 endoscopic datasets, designed to evaluate diagnostic agents across a comprehensive diagnostic chain, from fine-grained visual perception to high-level diagnostic reasoning. Experiments show that EndoCogniAgent achieves 85.23\% average accuracy on perception tasks and 71.13\% clinical acceptance rate on reasoning tasks, with ablation analysis confirming that self-consistency validation and episodic state maintenance are individually critical to these gains.
arXiv:2605.18671v1 Announce Type: new
Abstract: Reconnections and rotations of lines are dual descriptions of the same saddle-shaped spacetime surface. We show that a reconnection between two line occurring over time is a single line that rotates over space progression. Both rotating lines and reconnections possess the same saddle shape sheet geometry in four-dimensional space-time, with different orientations. Cyclic precessing lines occurring over time are arrays of reconnections occurring spatially. We show that a magnetic reconnection occurring over time can be seen as a single continuous line vector potential rotating spatially, where the full evolution traces a saddle shape surface. Finally, we show that a single tilted spatiotemporal optical vortex precesses with spatial progression, and as a result can be seen as two vortices reconnecting. Given the unique spatiotemporal evolution, we also analyzed the relativistic angular momentum of these electromagnetic fields.
arXiv:2510.10930v3 Announce Type: replace
Abstract: Reasoning is not just about solving problems -- it is also about evaluating which problems are worth solving at all. Evaluations of artificial intelligence (AI) systems primarily focused on problem solving, historically by studying how models play games such as chess and Go. In this paper, we advocate for a new paradigm that assesses AI systems' evaluation of games. First, we introduce a formalism for evaluating such evaluations. We then leverage a large-scale dataset of over 100 novel board games and over 450 human judgments to compare evaluations produced by modern language and reasoning models against those of people and symbolic computational agents. We consider two kinds of evaluative queries: assessing the payoff (or fairness) and the funness of games. These queries span two dimensions relevant to the design of evaluations of AI evaluations: how complex a query is to compute and how difficult a query is to quantify. Our results show that reasoning models are generally more aligned to people in their evaluations of games than non-reasoning language models. However, we observe a non-monotonic relationship: as models get closer to game-theoretic optimal, their fit to human data weakens. We also observe more "jaggedness" across models for assessing funness, in line with the greater difficulty of quantifying this query. Across queries and games, reasoning models show highly variable and unpredictable resource usage when assessing queries, pointing to the importance of imbuing more resource-rational meta-reasoning in language and reasoning models.
arXiv:2511.16361v3 Announce Type: replace
Abstract: Recent guided depth super-resolution methods are premised on the assumption of strict spatial alignment between depth and RGB, achieving high-quality depth reconstruction. However, in real-world scenarios, the acquisition of strictly aligned RGB-D is hindered by inherent hardware limitations (e.g., physically separate RGB-D sensors) and unavoidable calibration drift induced by mechanical vibrations or temperature variations. Consequently, existing approaches often suffer inevitable performance degradation when applied to misaligned real-world scenes. In this paper, we propose the Multi-Order Matching Network (MOMNet), a novel alignment-free framework that adaptively retrieves and selects the most relevant information from misaligned RGB. Specifically, our method begins with a multi-order matching mechanism, which jointly performs zero-order, first-order, and second-order matching to comprehensively identify RGB information consistent with depth across multi-order feature spaces. To effectively integrate the retrieved RGB and depth, we further introduce a multi-order aggregation composed of multiple structure detectors. This strategy uses multi-order priors as prompts to facilitate the selective feature transfer from RGB to depth. Extensive experiments demonstrate that MOMNet achieves superior performance and generalization across both unaligned and aligned datasets.
arXiv:2605.18672v1 Announce Type: new
Abstract: This position paper argues that enforcing LLM agent safety within a single abstraction layer is not merely suboptimal but categorically insufficient for deployed LLM agents -- a structural consequence of how agent execution works, not a contingent limitation of current systems. The three dimensions that jointly constitute safe operation -- semantic intent and policy compliance, environmental validity, and dynamical feasibility -- each depend on a strictly distinct set of information that becomes available at different stages of execution. No single guardrail can certify all three. We argue that the community must respond with a contract-based architecture in which each safety dimension is enforced by an independently certified layer whose probabilistic guarantee satisfies the next layer's assumption. We sketch such an architecture and derive the compositional system-level safety bounds it admits via the chain rule of probability. Three open problems stand between this and a deployable standard: bound estimation from non-i.i.d.\ traces, graceful degradation of contracts under deployment drift, and extension to multi-agent settings -- the most important unfinished business in LLM agent runtime assurance.
arXiv:2603.02531v2 Announce Type: replace
Abstract: Classifier-Free Guidance (CFG) improves sample quality in diffusion models, but its dual-pass inference and reliance on null-condition training limit its use in few-step regimes. Attention-space guidance has emerged as a complementary paradigm that addresses this gap, yet why prior sparse-vs-dense attention guidance works remains elusive. We address this by analyzing attention extrapolation through Modern Hopfield dynamics, proving two directional properties of the sparse-dense discrepancy under shared conditioning that together certify it as a directionally consistent acceleration signal. Building on this, we propose Geometry-Aware Attention Guidance (GAG), a training-free, plug-and-play extrapolation rule that decomposes the discrepancy into parallel and orthogonal components relative to the retrieval direction, amplifying the convergence-aligned component while suppressing off-manifold noise; stability follows from a weak contraction property. We further provide an interpretation of this extrapolation as first-order Anderson Acceleration in attention space, offering a unified perspective on attention extrapolation methods. GAG is a universal method that generalizes across architectures (UNet, MMDiT) and sampling regimes (multi-step, few-step), consistently improving generation quality on diverse backbones, including FLUX.1, the recent FLUX.2, and Qwen-Image, with minimal computational overhead.
arXiv:2603.03190v3 Announce Type: replace
Abstract: During music listening, cortical activity encodes both acoustic and expectation-related information. Prior work has shown that ANN representations resemble cortical representations and can serve as supervisory signals for EEG recognition. Here we show that distinguishing acoustic and expectation-related ANN representations as teacher targets improves EEG-based music identification. Models pretrained to predict either representation outperform non-pretrained baselines, and combining them yields complementary gains that exceed strong seed ensembles formed by varying random initializations. These findings show that teacher representation type shapes downstream performance and that representation learning can be guided by neural encoding. This work points toward advances in predictive music cognition and neural decoding. Our expectation representation, computed directly from raw signals without manual labels, reflects predictive structure beyond onset or pitch, enabling investigation of multilayer predictive encoding across diverse stimuli. Its scalability to large, diverse datasets further suggests potential for developing general-purpose EEG models grounded in cortical encoding principles.
arXiv:2603.03538v3 Announce Type: replace
Abstract: Large Language Models (LLMs) with chain-of-thought generation have demonstrated great potential for solving complex reasoning and planning tasks. However, the output of current LLMs is not fully reliable and needs careful verification. Even if LLMs get more accurate over time, learned verifiers can help increase trust, enforce safety constraints, and ensure alignment with personal preferences. A major challenge in learning verifiers, however, especially when their output will be used by the generator to improve its reasoning, is that the feedback loop between generator and verifier may produce substantial distribution shift. Motivated by this challenge, we propose an online learning framework for learning chain-of-thought verifiers that, given a problem and a sequence of reasoning steps, check the correctness of the solution. Highlighting the asymmetric role of soundness errors (failure in catching errors in a reasoning trace) and completeness errors (flagging correct reasoning steps as wrong), we introduce novel extensions of the Littlestone dimension which tightly characterize the mistake bounds for learning a verifier in the realizable setting. We provide optimal algorithms for finding the Pareto-frontier (the smallest total number of mistakes given a budget of soundness mistakes) as well as for minimizing a linear combination of asymmetric costs. We further show how our learned verifiers can be used to boost the accuracy of a collection of weak generators, and enable generation of proofs beyond what they were initially trained on. With the mild assumption that one of the generators can generate the next reasoning step correctly with some minimal probability, we show how to learn a strong generator with small error and abstention rates.
arXiv:2603.04727v2 Announce Type: replace
Abstract: Multimodal large language models (MLLMs) have demonstrated impressive general competence in video understanding, yet their reliability for real-world Video Anomaly Detection (VAD) remains largely unexplored. Unlike conventional pipelines relying on reconstruction or pose-based cues, MLLMs enable a paradigm shift: treating anomaly detection as a language-guided reasoning task. In this work, we systematically evaluate state-of-the-art MLLMs on the ShanghaiTech and CHAD benchmarks by reformulating VAD as a binary classification task under weak temporal supervision. We investigate how prompt specificity and temporal window lengths (1s--3s) influence performance, focusing on the precision--recall trade-off. Our findings reveal a pronounced conservative bias in zero-shot settings; while models exhibit high confidence, they disproportionately favor the 'normal' class, resulting in high precision but a recall collapse that limits practical utility. We demonstrate that class-specific instructions can significantly shift this decision boundary, improving the peak F1-score on ShanghaiTech from 0.09 to 0.64, yet recall remains a critical bottleneck. These results highlight a significant performance gap for MLLMs in noisy environments and provide a foundation for future work in recall-oriented prompting and model calibration for open-world surveillance, which demands complex video understanding and reasoning.