arXiv:2605.30702v1 Announce Type: new
Abstract: Active mid-infrared (MIR) imagers capable of retrieving three-dimensional (3D) structure and reflectivity information are highly attractive in a wide range of biomedical and industrial applications. However, the infrared 3D imaging at low-light levels is still challenging due to the deficiency of sensitive and fast MIR sensors. Here we propose and implement a MIR time-of-flight imaging system that operates at single-photon sensitivity and femtosecond timing resolution. Specifically, back-scattered infrared photons from a scene are optically gated by delay-controlled ultrashort pump pulses through nonlinear frequency upconversion. The upconverted images with time stamps are then recorded by a silicon camera to facilitate the 3D reconstruction with high lateral and depth resolutions. Moreover, an effective numerical denoiser based on spatiotemporal correlation allows us to reveal the object profile and reflectivity under photon-starving conditions with a detected flux below 0.05 photons/pixel/second. The presented MIR 3D imager features with high detection sensitivity, precise timing resolution, and wide-field operation, which may open new possibilities in life and material sciences.
Science Journals
arXiv:2605.31083v1 Announce Type: new
Abstract: Extending megaelectronvolt ultrafast electron diffraction (MeV UED) into the attosecond regime is essential for resolving intrinsic structural dynamics, yet requires simultaneously controlling electron-pulse duration and arrival-time stability. Here, we propose a generalized harmonic laser-electron interaction that extends beam modulation into a continuous harmonic regime. We demonstrate that highly detuned, non-integer harmonic modulation via a single-period undulator achieves stronger coupling efficiency than conventional integer-harmonic resonance. Driven by a mid-infrared seed laser whose wavelength is a small fraction of the nominal resonant wavelength, this mechanism enables effective longitudinal phase space manipulation. It facilitates attosecond compression with minimal laser-induced energy spread, preserving the beam quality required for high-fidelity diffraction. Furthermore, deriving both the modulation and experimental pump lasers from a common source intrinsically locks their relative timing. Simulations demonstrate 680-as pulse durations and 470-as arrival-time jitter, establishing a viable route to attosecond MeV UED for resolving coupled electron-nuclear dynamics.
arXiv:2605.31080v1 Announce Type: new
Abstract: Blind and low-vision (BLV) audiences remain underserved by visual art descriptions, particularly across languages and in museum settings where privacy and intellectual-property constraints may favour small on-premise vision-language models (VLMs). This pilot study investigates curator-guided multilingual art description with Qwen2.5-VL-3B-Instruct for German, Romanian, and Serbian. We construct a parallel BLV-oriented caption corpus from artwork images and metadata, and compare language-specific LoRA adapters with a single multilingual adapter under a fixed backbone and training budget. Evaluation combines automatic lexical and embedding-based metrics with an LLM-as-Judge protocol calibrated against a small Romanian BLV pilot study. Under our pilot setup, language-specific adapters show more stable controllability and visually grounded description quality for Romanian and Serbian, while multilingual adaptation remains competitive in German. We frame these findings as deployment-oriented evidence for small on-premise VLMs, and highlight the need for larger BLV user studies and broader language coverage before drawing general conclusions about multilingual accessibility.
arXiv:2605.29317v2 Announce Type: replace
Abstract: Parameter-efficient fine-tuning(PEFT) has largely focused on LoRA and its accuracy-oriented variants, leaving the original goal of reducing trainable parameters has receivedcomparatively little attention. We introduce FoRA, which revisits this goal by reducing the number of adapted layers rather than adapter rank. FoRA selects task-informative layers via a single-pass diagonal Fisher score (under 1% of training cost) and trains the LoRA down-projection at selected layers on the Stiefel manifold, preserving column orthonormality and effective rank. FoRA consistently outperforms LoRA and DoRA at half their parameter budget, and falls within 0.7-0.8 accuracy points of AdaLoRA at one-quarter its parameter count, across five LLaMA-family backbones. Cross-architecture experiments on twelve backbones from the LLaMA, Qwen3, and Gemma families confirm consistent gains from 270M to 32B parameters. The two components combine super-additively: Fisher selection alone matches rank reduction at the same budget, while the Stiefel constraint provides the decisive additional gain.
arXiv:2605.31079v1 Announce Type: new
Abstract: This study compares two methodological approaches for predicting, at a given site, threshold exceedances of atmospheric variables such as temperature and wind speed: (i) direct probabilistic methods, which treat exceedance as a binary classification problem, and (ii) full distribution probabilistic methods, which model the complete conditional probability law of the target variable. Using theoretical analysis and numerical simulations on a toy model, alongside real-world data from the MeteoNet dataset (2016--2018) for southeastern France, we demonstrate that the full distribution approach consistently outperforms the direct method for rare, extreme events. This advantage arises because the full distribution approach effectively learns the parameters of the conditional distribution from moderate and mild intensity events, thereby achieving better calibration and discrimination in the tails. We find that the specific parametric shape of the chosen distribution plays a secondary role compared to accurately capturing predictable shifts in its bulk properties (i.e., mean and variance). This empirical indistinguishability is also informative about the physical mechanics driving atmospheric extremes, suggesting that extreme exceedances are primarily driven by significant conditional displacements of the entire distribution rather than by unpredictable, fat-tailed anomalies within a static climatology. Our results are validated for both strong surface wind speeds and intense hourly rainfall, with performance evaluated using proper scoring rules (Brier score, logarithmic score) and deterministic skill scores (Peirce Skill Score, CSI, HSS). These findings highlight the critical importance of modeling the full probability distribution for rare-event forecasting and provide practical guidance for improving extreme weather prediction in operational meteorology.
arXiv:2506.22304v3 Announce Type: replace
Abstract: Continuous Normalizing Flows (CNFs) enable elegant generative modeling but remain bottlenecked by their iterative nature requiring costly sampling and lacking interpretability of the intermediate states. Recent approaches accelerate sampling by straightening trajectories or distilling endpoints, yet they treat the original generative process as a black box, discarding the teacher's intermediate dynamics. We propose a fundamentally different perspective: globally linearizing flow dynamics via Koopman theory to achieve trajectory-preserving linearization. By lifting a pre-trained Conditional Flow Matching (CFM) model into a higher-dimensional Koopman space, we represent its evolution with a single linear operator. Crucially, unlike boundary-only distillation, our method enforces infinitesimal consistency with the teacher's vector field along the full generative path. We derive a practical, simulation-free training objective that ensures this global alignment and yields two key benefits. First, sampling becomes one-step and parallelizable. Second, because the linearization is faithful to the dynamics, the Koopman operator provides unique insights on the generation. We demonstrate that this structure enables novel applications unavailable in prior approaches, including discovery of semantically coherent editing directions, inversion with a teacher-aligned linear operator and class-conditional spectral signatures. Empirically, our approach achieves competitive sample quality, while enabling spectral analysis and control of the entire trajectories of generative flows.
arXiv:2605.31075v1 Announce Type: new
Abstract: Long-term memory is essential for multimodal agents to build coherent experience, accumulate world knowledge, and achieve continual learning. However, constructing effective memory goes beyond memory module design and basic requirements such as accuracy and fidelity; the key challenge lies in determining what to memorize. Multimodal agents, such as embodied agents, continuously perceive, reason, and act in real or virtual environments, receiving an unbounded stream of multimodal observations. From this combinatorial explosion of information, an agent must selectively retain content that is relevant to its role in the environment and valuable for future tasks. To bridge this gap, we frame memory generation as a learnable memorization policy and introduce TaskMem (Task-focused Memorization Policy Learning), a reinforcement-learning-based framework that enables the policy to dynamically adjust its focus to the demands of real tasks encountered in the environment. TaskMem adopts a two-phase training paradigm: Phase One learns how to memorize by optimizing memory quality under fundamental fidelity requirements; Phase Two occurs after deployment, where the agent learns what to memorize by tuning an adapter on its base MLLM, using recent environment tasks to define a reward model that guides the memorization policy toward task-relevant content. To evaluate our approach, we reformulate VideoMME, EgoLife, and EgoTempo into streaming benchmarks that simulate a realistic setting in which an agent processes streaming observations and handles tasks arriving online. To isolate memory assessment, the questions must be answered using only the agent's memory, without access to raw video. Built on Qwen3-VL-30B-A3B, TaskMem improves VQA accuracy by 6.3%, 7.0%, and 5.3% on these benchmarks, respectively.
arXiv:2605.30703v1 Announce Type: new
Abstract: Single-pixel cameras have recently emerged as promising alternatives to multi-pixel sensors due to reduced costs and superior durability, which are particularly attractive for mid-infrared (MIR) imaging pertinent to applications including industry inspection and biomedical diagnosis. To date, MIR single-pixel photon-sparse imaging has yet been realized, which urgently calls for high-sensitivity optical detectors and high-fidelity spatial modulators. Here, we demonstrate a MIR single-photon computational imaging with a single-element silicon detector. The underlying methodology relies on nonlinear structured detection, where the encoded time-varying pump patterns are optically imprinted onto the MIR object image through sum-frequency generation. Simultaneously, the MIR radiation is spectrally translated into the visible region, thus permitting infrared single-photon upconversion detection. Then, the use of advanced algorithms of compress sensing and deep learning allows us to reconstruct MIR images under sub-Nyquist sampling and photon-starving illumination. The presented paradigm of single-pixel upconversion imaging is featured with single-pixel simplicity, single-photon sensitivity, and room-temperature operation, which would establish a new path for sensitive imaging at longer infrared wavelengths or terahertz frequencies, where high-sensitivity photon counters and high-fidelity spatial modulators are typically hard to access.
arXiv:2506.12060v2 Announce Type: replace
Abstract: Cybersecurity organizations are adapting to GenAI integration through modified frameworks and hybrid operational processes, with success influenced by existing security maturity, regulatory requirements, and investments in human capital and infrastructure. This qualitative research employs systematic document analysis and comparative case study methodology to examine how 25 studies from 2022 to 2025 document organizational adaptation of threat modeling frameworks, revealing a shift away from traditional signature-based systems toward AI-capable frameworks across three primary patterns: LLM integration for security applications, GenAI frameworks for risk detection and response automation, and AI/ML integration for threat hunting and matching. Organizations with mature infrastructures, particularly in finance and critical infrastructure, demonstrate higher readiness through structured governance, dedicated AI teams, and robust incident response processes, with central banks and financial institutions leading adaptation efforts under regulatory pressure. Successful integration requires human oversight of automated systems, attention to data quality and explainability, and sector-specific governance, though ongoing difficulties with privacy protection, bias reduction, personnel training, and adversarial defense persist. Notable imbalances between offensive and defensive GenAI capabilities create strategic concerns for security planning. The findings offer actionable insights for cybersecurity professionals and underscore the need for adaptive approaches, ethical frameworks, and staff development when managing AI-enhanced threats.
arXiv:2509.06856v2 Announce Type: replace-cross
Abstract: We propose a novel randomized framework for the estimation problem of large-scale linear statistical models, namely Sequential Least-Squares Estimators with Fast Randomized Sketching (SLSE-FRS), which integrates Sketch-and-Solve and Iterative-Sketching methods for the first time. By iteratively constructing and solving sketched least-squares (LS) subproblems with increasing sketch sizes to achieve better precisions, SLSE-FRS gradually refines the estimators of the true parameter vector, ultimately producing high-precision estimators. We analyze the convergence properties of SLSE-FRS, and provide its efficient implementation. Numerical experiments show that SLSE-FRS outperforms the state-of-the-art methods, namely the Preconditioned Conjugate Gradient (PCG) method, and the Iterative Double Sketching (IDS) method.
arXiv:2605.30885v1 Announce Type: new
Abstract: This paper introduces a novel framework for polar codes, designed for flexible Incremental Redundancy Hybrid Automatic Repeat Request (IR-HARQ). By generalizing the decoding order beyond the standard 1$\to$N sequence, we enable a capacity-aware scheduling strategy that prioritizes the decoding of reliable subblocks. The framework integrates nested parity-check polar construction and reverse bit-mapping to support continuous and arbitrary transmission lengths $E \in [N_{\min}, N_{\max}]$. Simulation results show that the proposed rateless codes match the coding gain of independently optimized fixed-rate codes across the entire range of rates and lengths. With a validated hardware implementation, this work provides a practical solution for next-generation wireless data channels.
arXiv:2505.03416v2 Announce Type: replace
Abstract: Fabry-Perot cavities are widely used in current gravitational-wave detectors. In particular, they play a key role in frequency-dependent squeezing systems, enabling broadband quantum noise reduction. However, their ability to precisely control squeezing properties may be insufficient for the next generation of detectors. In this context, theoretical studies on linear three-mirror cavities have revealed promising features, such as resonance peak splitting and their equivalence to two-mirror cavities with variable finesse. In this paper, we report experimental observations of both the resonant doublet and variable finesse using a meter-scale implementation of a linear three-mirror cavity.
arXiv:2605.31096v1 Announce Type: new
Abstract: While visually grounded Chain-of-Thought (CoT) has emerged as a promising paradigm to enhance fine-grained perception in multimodal large language models (MLLMs), its efficacy during the inference phase remains underexplored. In this work, we empirically find that mandating explicit object boxes in visually grounded CoT during inference often degrades performance compared to standard textual CoT, which reasons without explicit visual grounding. We hypothesize that the visual localization capability can be internalized into the textual CoT and that the mandatory explicit grounding introduces unnecessary interference with the model's primary objective of answer prediction. To address this problem, we propose Internalizing Visually Grounded Reasoning (\textbf{iVGR}), a novel reinforcement learning framework that transfers localization capabilities into the textual reasoning process. We employ a dual-stream training strategy, where a textual stream is aligned with a high-quality visually grounded stream via a proposed consistency reward, enabling the model to localize accurately without explicit grounding during inference. Extensive experiments demonstrate that our method significantly outperforms existing baselines on fine-grained benchmarks, while maintaining the flexibility to support tool-assisted inference workflows.
arXiv:2510.15340v3 Announce Type: replace-cross
Abstract: State preparation is a cornerstone of quantum technologies, underpinning applications in computation, communication, and sensing. Its importance becomes even more pronounced in non-Markovian open quantum systems, where environmental memory and model uncertainties pose significant challenges to achieving high-fidelity control. Invariant-based inverse engineering provides a principled framework for synthesizing analytic control fields, yet existing parameterizations often lead to experimentally infeasible, singular pulses and are limited to simplified noise models such as those of Lindblad form. Here, we introduce a generalized invariant-based protocol for finite-dimensional state preparation under arbitrary noise conditions. We transform the finite-dimensional control problem into the equivalent problem for a single-qubit, by restricting the dynamics to a designed SU(2) subspace. The control protocol then proceeds in two-stages: first, we construct a family of bounded pulses that achieve perfect state preparation in a closed system; second, we identify the optimal member of this family that minimizes the effect of noise. The framework accommodates both (i) characterized noise, enabling noise-aware control synthesis, and (ii) uncharacterized noise, where a noise-agnostic variant preserves robustness without requiring a master-equation description. Numerical simulations demonstrate high-fidelity state preparation across diverse targets while producing smooth, hardware-feasible control fields. This singularity-free framework extends invariant-based control to realistic open-system regimes, providing a versatile route toward robust quantum state engineering on NISQ hardware and other platforms exhibiting non-Markovian dynamics.
arXiv:2605.22967v2 Announce Type: replace
Abstract: When Masked Diffusion Models (MDMs) generate sequences through iterative refinement, the rich internal computation over masked positions is discarded, forcing every subsequent refinement step to recompute the valuable internal information stored as model representations. To avoid a hard reset between denoising rounds, we propose Learned Relay Representations (Relay), a method that allows MDMs to be forward-thinking when denoising by explicitly learning how to propagate latent information for the benefit of future denoising steps. Relay introduces a differentiable per-token channel that passes information between forward passes and is trained via truncated backpropagation through time (BPTT). We show that this framework can be scaled to state-of-the-art Diffusion Language Models (DLMs), and is seamlessly compatible with techniques like block diffusion and KV caching. We first provide a thorough justification of the design choices in Relay on a challenging Sudoku-based planning task. We then scale Relay to Fast-dLLM v2, a state-of-the-art DLM, outperforming standard supervised finetuning on coding tasks while reducing inference latency by up to 32%. Our empirical results demonstrate that state-of-the-art DLMs can be explicitly trained to relay latent information forward across decoding steps, advancing the performance-latency Pareto frontier. We provide code for all our experiments.
arXiv:2605.26711v2 Announce Type: replace
Abstract: We construct a binary mixed-regime process with one deterministic textual regime and one random regime governed by an unobserved latent state. Even an ideal infinite-capacity sequence predictor that exactly recovers the text-only marginal law can become overconfident when the observed prefix is compatible with the wrong latent regime. The resulting entropy difference is not an ordinary optimization error; it is a sufficiency gap caused by marginalization over an unobserved state. We then formalize retrieval, tool use, and external grounding through an auxiliary binary signal with fidelity $\gamma \in [1/2,1]$. The resulting Bayesian update yields a contextual dominance threshold: a corrective signal reverses the posterior odds induced by the textual history exactly when its fidelity exceeds the text-only posterior weight assigned to the misleading regime. This threshold reduces, but does not generally eliminate, the sufficiency gap; complete closure requires perfect revelation of the relevant latent state or an equivalent verification mechanism. The analysis clarifies why temperature scaling cannot restore missing context, why grounding mechanisms must be both informative and learnably usable by the model, and why autonomous sequence models require structurally decoupled observers or verifiers in high-stakes domains.
arXiv:2605.30711v1 Announce Type: new
Abstract: Agentic LLMs must continuously decide whether newly extracted facts should be added, merged with existing memories, or ignored, yet prior work has focused more on retrieval and storage than on principled write-side control. We frame memory evolution as a novelty-detection problem and propose SAGE, a Spherical Adaptive Gate for memory Evolution that scores candidate facts with a von Mises-Fisher-based density estimator over memory embeddings and routes them with an adaptive threshold that tracks memory-store geometry. SAGE resolves clearly novel facts as ADD, clearly redundant facts as NOOP, and sends only uncertain cases to an LLM merge step, reducing expensive write-time reasoning. On LoCoMo, SAGE achieves the best average token-F1 against Mem0 on all seven open-weight backbone comparisons, while on GPT-4o-mini it reduces add-phase API cost by 3.4$\times$ and add-phase latency by 2.5$\times$ with only a small average judge-score gap. As a drop-in binary gate for A-Mem, SAGE skips roughly 16-18% of LLM calls across five models with minimal quality change on open-weight backbones. These results suggest that novelty-aware write control is a practical lever for improving both memory quality and system efficiency in long-term agentic memory.
arXiv:2605.31126v1 Announce Type: new
Abstract: Can a language model improve from plain text sampled from itself, with no prompts, no teacher, no verifier, and no reward model? Yes, but only when the synthetic corpus is compatible with the student, a relational property of the source-student pair rather than an intrinsic property of the data. We call this the latent capability resurfacing hypothesis: weak self-training can amplify capabilities already present in the pretrained model, but only under this compatibility condition. We study this in the minimal setting of prompt-free unconditional self-training, where base language models are fine-tuned on text generated from the BOS token alone, with no task specification or external supervision. We report three findings. First, synthetic utility is relational rather than intrinsic: self-generated data is the most effective source, same-lineage transfer outperforms stronger but differently trained sources, and cross-family transfer is substantially weaker. Second, common intrinsic proxies fail: neither benchmark-level semantic similarity nor average per-token likelihood under the student predicts which corpora help. Third, this regime produces a surprising byproduct. In controlled Pythia experiments, capability and verbatim memorization decouple: benchmark utility is preserved or improved while held-out exact-match extraction drops by over 95 percent, with no forget set, privacy objective, or targeted unlearning. Together, these results suggest that prompt-free self-training works by amplifying what the student already knows, not by importing structure from the data. They also reveal a regime in which capability and verbatim memorization can be separated without any explicit unlearning objective.
arXiv:2605.26371v2 Announce Type: replace
Abstract: Hierarchical Reinforcement Learning (HRL) promises to solve long-horizon Reinforcement Learning (RL) tasks more efficiently than non-hierarchical counterparts by discovering and reusing temporally-extended skills. However, obtaining skills that are actually reusable remains an open challenge. Towards this end, we focus on abstractions that exploit the intuition of local dynamics: local transitions in different global contexts require similar kinds of action sequences. By aligning these contexts with the action sequences they require, we are able to learn which skills to reuse and where to reuse them. In principle, this information should benefit many HRL algorithms, where high-level policies have to reason about the low-level skills they use. The resulting algorithm CARL (Contrastive Action-based Representations for Reusable Local Control) shows both qualitative clustering of meaningful skills in complex humanoid environments and improved downstream performance on the OGBench benchmark when integrated with HIQL.
arXiv:2605.30713v1 Announce Type: new
Abstract: Test-time compute (TTC) strategies have emerged as a lightweight approach to boost reasoning in large language models (LLMs). However, their application and benefits for vision-language models (VLMs) remain underexplored. We present a systematic study of TTC across seven VLMs and six benchmarks, specifically analyzing feature-based scoring and majority voting methods. We find that feature heuristics fail and voting yields only modest gains in single-model settings. We theoretically show that this limitation stems from a lack of prediction diversity: when outputs are highly correlated, voting provides little benefit. In contrast, multi-model ensembles offer richer diversity, yet standard majority voting fails to account for varying model capabilities. To address this, we propose Entropy-based TTC (ETTC), which selects the most confident prediction based on predictive entropy. Our method reduces to majority voting in the single-model case, but in model ensembles, it leverages confidence disparities to prioritize stronger models. We prove that ETTC outperforms majority voting under mild assumptions and empirically demonstrate that it consistently surpasses both voting and the best individual model. Crucially, our results show that smaller models can synergistically enhance larger ones, unlocking ensembling gains not achievable with standard strategies.
arXiv:2605.31243v1 Announce Type: new
Abstract: This study introduces a novel Air Traffic Control (ATC) concept to support self-separation between vehicles in Urban Air Mobility (UAM) corridors. Our proposed scheme involves sharing intended arrival schedules at Constrained Waypoints (CWPs) among UAM operators. We propose two approaches to assist the arrival scheduling at CWPs by computing the minimum arrival time gap necessary for each pair of vehicles to ensure their safety throughout the flights within the corridor. The first approach considers the minimum separation distance required by the Near Mid-Air-Collision (NMAC) avoidance rules, while the second one is based on the Responsibility-Sensitive Safety (RSS) rules. We demonstrate that the NMAC-rule-based approach can effectively prevent collisions in normal circumstances, where the vehicles adhere to the speed limits of the corridor. However, this approach does not guarantee safety if vehicles exceed the speed limits. Conversely, while the RSS-rule-based approach ensures collision prevention during emergencies when vehicles exceed speed limits, it may require larger arrival time gaps under normal circumstances, which may lead to reduced traffic flow. Our results are confirmed through numerical simulations.
arXiv:2605.31270v1 Announce Type: new
Abstract: This paper studies the control of discrete-time linear fractional-order networks, a flexible modeling framework for systems with long-range memory such as power grids, biological networks, and neuronal circuits. In contrast to the common view that fractional exponents (time-scales) are fixed parameters, we show that they can be systematically steered, together with the network coupling matrix, by appropriately designed input sequences. We first derive algebraic conditions under which the coupling matrix and the vector of fractional exponents of a given network can be reconfigured to desired values, and we characterize how truncating the infinite-memory term impacts the resulting dynamics. Building on these results, we construct an equivalent linear representation that isolates the contribution of memory, and we introduce a fractional reachability matrix that provides explicit conditions for jointly steering both network parameters and state in a finite number of steps. To address practical implementations, we further formulate an energy-constrained steering problem that incorporates actuator bounds and finite-memory approximations as a quadratic program. The framework is illustrated on low-dimensional toy examples, on larger networks with Erdos-Renyi, Barabasi-Albert, and Watts-Strogatz topologies, and on a brain network model inferred from electrocorticography recordings of an epilepsy patient, where we showcase transitions between pre-seizure and seizure configurations.
arXiv:2505.20795v2 Announce Type: replace
Abstract: Recent robot learning methods commonly rely on imitation learning from massive robotic dataset collected with teleoperation. When facing a new task, such methods generally require collecting a set of new teleoperation data and finetuning the policy. Furthermore, the teleoperation data collection pipeline is also tedious and expensive. Instead, human is able to efficiently learn new tasks by just watching others do. In this paper, we introduce a novel two-stage framework that utilizes human demonstrations to learn a generalizable robot policy. Such policy can directly take human demonstration video as a prompt and perform new tasks without any new teleoperation data and model finetuning at all. In the first stage, we train video generation model that captures a joint representation for both the human and robot demonstration video data using cross-prediction. In the second stage, we fuse the learned representation with a shared action space between human and robot using a novel prototypical contrastive loss. Empirical evaluations on real-world dexterous manipulation tasks show the effectiveness and generalization capabilities of our proposed method.
arXiv:2605.22737v3 Announce Type: replace
Abstract: Distillation attacks create a deployment trade-off for model providers: the same outputs that make a model more useful can also make it easier to imitate. We study this trade-off through a minimax game between a utility-constrained teacher and an adaptive student. Our framework yields tractable one-sided response rules: an adaptive evaluation rule in which the student reweights high-value examples, and a teacher-side defense template that suppresses outputs most useful for distillation. From a cheap proxy for example value, we derive Product-of-Experts (PoE), a simple forward-pass-only defense that combines the teacher with a proxy student during generation. Empirically, adaptive evaluation reveals a large passive--adaptive gap: on state-of-the-art defenses, adaptive students recover substantially more capability than passive evaluation suggests on GSM8K and MATH. Under this stronger evaluation, the apparent robustness gap between expensive defenses and PoE narrows considerably, while PoE remains substantially cheaper and preserves higher-quality reasoning traces. Overall, our results suggest that strong distillation remains difficult to stop, and that progress on antidistillation should be judged against adaptive students rather than passive ones. Our code is available at: https://github.com/ysfalh/distillation-game.
arXiv:2605.31074v1 Announce Type: new
Abstract: Many advancements in optics have relied on the tight-binding approximation, which simplifies the description and prediction of complex system behaviors. This approximation describes the dynamics of the total light field by examining the coupling between the guided modes of individual single-mode substructures -- also known as coupled mode theory. However, the underlying assumption, that the guided modes of individual waveguides form an orthogonal basis, breaks down when waveguides are brought into close proximity or when larger arrays are considered. In this work, we systematically analyze the consequences of this non-orthogonality and show that it leads to a generalized eigenvalue problem involving an overlap matrix, causing a fundamental mismatch between the standard TB model and solutions of the paraxial wave equation. To resolve this issue, we introduce a modified TB framework based on the L\"owdin orthogonalization, which constructs an orthonormal basis from the non-orthogonal guided modes while minimally altering their physical shape and preserving their symmetry properties. The resulting L\"owdin-TB method restores the standard eigenvalue problem and yields excellent agreement with exact beam propagation simulations across a wide range of system sizes and waveguide separations. Furthermore, it captures important physical effects, such as enhanced long-range coupling and nontrivial hopping phases, that are absent in the standard approach.