arXiv:2605.31108v1 Announce Type: new
Abstract: In this work we introduce a novel approach to domain incremental learning, adapting models over time to evolving, non-stationary data. In contrast to other works, we do not attempt to avoid catastrophic forgetting, but rather allow it and exploit it. Our model combines a main task head with a self-supervised masked autoencoder (MAE) head. We then learn domain-specific LoRA adapters during incremental training. Each adapter specializes to its domain, naturally inducing forgetting on other domains in both heads. At inference, we perform online test-time training on the self-supervised MAE head to identify which LoRAs best matches the current input, so the model can `remember' the domain again. Our scheme is especially well-suited to real-world streaming data, such as video, where consecutive samples are highly correlated and domain shifts are gradual. We demonstrate our method on domain-incremental action recognition and semantic segmentation tasks.
Science Journals
arXiv:2605.30406v1 Announce Type: new
Abstract: Recent research demonstrating AI systems exhibiting deception and shutdown resistance suggests that AI loss of control (LOC) is an urgent policy concern , yet current literature focuses almost exclusively on alignment and prevention. To address this gap, this paper introduces a foundational framework and taxonomy for managing catastrophic AI LOC incidents. The taxonomy's first level distinguishes between scenarios where regaining control is 'extremely costly' versus 'impossible'. While impossible scenarios demand immediate resilience investments to fundamentally restrict an AI's attack surface , extremely costly scenarios require active incident management via Containment and Threat Neutralization. The framework further categorizes these manageable events into accidental LOC (requiring automated circuit-breaker responses) and adversarial LOC (requiring graduated escalatory measures). By mapping three severity classes to specific scenario matrices, this paper provides a concrete, proportional guide for managing unprecedented AI risks.
arXiv:2605.18861v3 Announce Type: replace
Abstract: We present a deep neural net-based region of interest detection method (DNN ROI) for signal processing in the liquid argon time projection chambers of the Short-Baseline Neutrino (SBN) Program, SBND and ICARUS. DNN ROI addresses limitations of the traditional wire-by-wire thresholding algorithm by leveraging the full two-dimensional detector readout and cross-plane matching information. To account for detector performance variations, we explore training with augmented samples. We find that DNN ROI outperforms the traditional method in both low-level ROI identification performance and high-level reconstruction metrics for high-energy cosmic and accelerator neutrino interaction products, while also being more robust against detector variations, with or without sample augmentation.
arXiv:2605.30613v1 Announce Type: new
Abstract: Over the past year, prompt caching in Large Language Models (LLMs) has become increasingly more popular across inference APIs. Prompt caching helps save precious compute resources and speeds up response times by reusing parts of the KV cache of a specific prompt for another request. However, many implementations of prompt caching are not secure against timing attacks or even basic metadata disclosure. Gu et al. (ICML 2025) develop a method to audit prompt caching in LLMs. This paper investigates whether OpenRouter's API gateway architecture introduces prompt caching vulnerabilities that bypass provider-level prompt cache isolation guarantees. Most LLM inference providers implement per-account or per-organization prompt caching to prevent data leaks, but does routing through OpenRouter with shared organizational credentials inadvertently create global cache sharing across all OpenRouter users?
arXiv:2605.30986v1 Announce Type: cross
Abstract: We theoretically investigate strong-coupling properties of an attractively interacting Fermi atomic gas, where the Cooper-pair formation occurs between atoms belonging to different dimensional bands. Including pairing fluctuations within the framework of the self-consistent $T$-matrix approximation (SCTMA), we examine how the BCS-type superfluid phase transition temperature $T_\mathrm{c}$ varies as one moves from the 3D-3D to the 2D-3D system, in the wide parameter region with respect to the strength of the pairing interaction. In the 2D-3D limit, we find that, while the mean-field BCS theory predicts $T_\mathrm{c}>0$ in the strong-coupling regime, $T_\mathrm{c}$ is remarkably suppressed down to zero by pairing fluctuations that are strongly enhanced by the mixed-dimensionality of the system. As the origin of this, we clarify that the lower-dimensional (2D) component dominates the superfluid instability, so that the vanishing $T_\mathrm{c}$ is the same phenomenon as that in the 2D-2D case. We also point out that this can already be seen in the mean-field level, when one examines the propagation of the Goldstone mode. On the other hand, we find that the pseudogap phenomenon, which is known as a precursor of Cooper-pair formation, exhibits a 3D character of the 2D-3D system. These results indicate that the dimensionality of a strongly interacting Fermi gas depends on what we observe.
arXiv:2510.10544v3 Announce Type: replace
Abstract: We derive a novel PAC-Bayesian generalization bound for reinforcement learning that explicitly accounts for Markov dependencies in the data, through the chain's mixing time. This contributes to overcoming challenges in obtaining generalization guarantees for reinforcement learning, where the sequential nature of data breaks the independence assumptions underlying classical bounds. The new bound provides non-vacuous certificates for modern off-policy algorithms such as Soft Actor-Critic. We demonstrate the practical utility of the bound through PB-SAC, a novel algorithm that optimizes the bound during training to guide exploration. Experiments across several continuous control tasks show that the proposed approach provides meaningful confidence certificates while maintaining competitive performance.
arXiv:2605.31141v1 Announce Type: new
Abstract: Spiking neural networks (SNNs) provide an efficient event-driven computing paradigm for bio-inspired interception tasks. However, most implementations rely on von Neumann digital computing platforms, where memory and computation bottlenecks limit energy efficiency. This work presents a compact and energy-efficient memristive neuromorphic accelerator for bio-inspired interception tasks. A novel one-transistor-one-resistor (1T1R) crossbar array is designed to emulate synaptic operations in the in-memory computing (IMC) domain, while circuit-level optimization mitigates membrane drift and improves integration fidelity. In addition, an integrate-and-fire (IF) neuron with separated input and membrane nodes is developed to improve inference robustness during array-interfaced operation. Implemented in the SkyWater SKY130 PDK, the proposed neuron achieves an energy consumption of 10.67 pJ/spike and an area of 906 um^2. System-level results show that the memristive IMC output closely matches the software SNN baseline, with a correlation coefficient of 0.9622, while achieving a 96% interception success rate. These results demonstrate the effectiveness of the proposed design for compact and reliable memristive SNN inference in bio-inspired interception tasks.
arXiv:2605.31574v1 Announce Type: new
Abstract: Radical Moral Disagreements (RMDs) are highly polarising topics that are increasingly censored in everyday life, with growing evidence suggesting that this polarisation carries measurable costs to public mental health. To address these challenges, some researchers have proposed Large Language Models (LLMs) as a means to support more democratic deliberation and better moral reasoning. Yet existing tools are poorly calibrated to help people navigate RMDs, because of their intense and divisive characteristics. This paper introduces CONSIDER, a prototype for a one-to-one AI tool for RMD navigation. Drawing on Mill's account of the epistemic value of disagreement, CONSIDER aims at value clarification through structured disagreement with an opposing LLM-generated opinion. We describe CONSIDER's design logic and analyse potential risks posed by such tools to guide future development.
arXiv:2605.31311v1 Announce Type: cross
Abstract: Networked AI systems increasingly rely on multiple agents that collaboratively learn and adapt models over communication networks. In such systems, bilevel formulations naturally arise in hyperparameter optimization, data cleaning, and meta-learning, but the repeated evaluation of gradients, Jacobians, and Hessians can impose a substantial computational burden on individual agents. To address this challenge, we propose Snapshot-SLDBO (S$^3$LDBO), an efficient single-loop decentralized bilevel optimization algorithm that enables agents to intermittently skip expensive derivative evaluations through a snapshot mechanism. This mechanism can be interpreted as an autonomous computation-adaptation strategy for networked AI, where agents selectively perform costly local updates while maintaining global collaborative learning. We establish the ergodic iteration complexity and the high probability nonergodic iteration complexity of the proposed algorithm within a deterministic setting. Experimental results on hyperparameter optimization with synthetic and MNIST datasets, data hyper-cleaning on Fashion-MNIST, and decentralized meta-learning on miniImageNet demonstrate that the proposed algorithm improves computational efficiency while maintaining competitive learning performance.
arXiv:2511.07965v3 Announce Type: replace-cross
Abstract: A least common ancestor (LCA) of two leaves in a directed acyclic graph (DAG) is a vertex that is an ancestor of both leaves and has no proper descendant that is also their common ancestor. LCAs capture hierarchical relationships in rooted trees and, more generally, in DAGs. In 1981, Aho et al. introduced the problem of determining whether a set of pairwise LCA constraints on a set $X$, of the form $(i,j)<(k,l)$ with $i,j,k,l\in X$, can be realized by a rooted tree whose leaf set is $X$, such that whenever $(i,j)<(k,l)$, the LCA of $i,j$ is a descendant of that of $k,l$. They also presented a polynomial-time algorithm, BUILD, to solve this problem. However, many such constraint systems cannot be realized by any tree, prompting the question of whether they can be realized by a more general DAG. We extend Aho et al.'s framework from trees to DAGs, providing both theoretical and algorithmic foundations for reasoning about LCA constraints in this broader setting. Given a collection $R$ of LCA constraints, we define its $+$-closure $R^+$, capturing additional LCA relations implied by $R$. Using $R^+$, we construct a canonical DAG $G_R$ and prove that $R$ is DAG-realizable if and only if it is realized by $G_R$. We further adapt this construction to phylogenetic networks, defining a canonical network $N_R$ and prove that it is regular, i.e., it coincides with the Hasse diagram of its underlying set system. Finally, we show that for any DAG-realizable $R$, its classical closure - comprising all LCA constraints that hold in every DAG realizing $R$ - coincides with its $+$-closure. All constructions are computable in polynomial time, and we provide explicit algorithms for each. All algorithms developed in this paper are implemented in the freely available Python package RealLCA.
arXiv:2605.30366v1 Announce Type: new
Abstract: Recent Singing Voice Synthesis (SVS) advances enable highly realistic but potentially malicious AI covers, making singing voice deepfake detection (SVDD) crucial. Self-Supervised Learning (SSL)-based detectors achieve state-of-the-art performance by fine-tuning speech SSL backbones to capture singing-specific spoof artifacts. Existing adversarial attacks often fail against SSL-SVDD, creating a false impression of inherent robustness. We reveal this stems from two challenges. First, at the objective level, attacks optimize cross-entropy on local surrogates, crossing surrogate-specific boundaries rather than suppressing shared spoof evidence. Second, at the method level, attacks follow the surrogate's dominant gradient direction. In SSL-SVDD, this aligns with fine-tuned artifact-sensitive directions, limiting transferability to unseen detectors - a geometric failure we term the Linearity Trap. To properly evaluate robustness, we propose MARS (Meta-Adversarial Regression of Semantics), a transfer-based black-box framework tailored to SSL-SVDD. Structurally, MARS shifts to hypothesis-evidence manipulation by constructing a natural semantic anchor from the pre-trained SSL space and an artifact anchor from the fine-tuned space. Algorithmically, MARS escapes the Linearity Trap via bi-level optimization: the inner stage induces tangential exploration, while the outer stage guides the audio toward the natural semantic manifold. Experiments on the CtrSVDD benchmark show MARS improves Attack Success Rate (ASR) in in-distribution transfer (13%), out-of-distribution transfer (10%), and cross-task evaluation (36%), highlighting the urgent need for robust SVDD systems.
arXiv:2507.17026v2 Announce Type: replace-cross
Abstract: The two-sample testing problem, a fundamental task in statistics and machine learning, seeks to determine whether two sets of samples, drawn from underlying distributions $p$ and $q$, are in fact identically distributed (i.e. whether $p=q$). A popular and intuitive approach is the classifier two-sample test (C2ST), where a classifier is trained to distinguish between samples from $p$ and $q$. Yet despite simplicity of the C2ST, its reliability hinges on access to a near-Bayes-optimal classifier, a requirement that is rarely met and difficult to verify. This raises a major open question: can a weak classifier still be useful for two-sample testing? We show that the answer is a definitive yes. Building on the work of Hu and Lei (2024), we analyze two conformal variants of the C2ST that convert the scores from any trained classifier -- even if weak, biased, or overfit -- into exact, finite-sample p-values. We establish two key theoretical properties of the conformal C2ST: (i) finite-sample Type-I error control, and (ii) non-trivial power that degrades gently in tandem with the error of the trained classifier. The upshot is that even poorly performing classifiers can yield powerful and reliable two-sample tests. This general framework finds a powerful application in Bayesian inference, particularly for validating Neural Posterior Estimation (NPE) models, where the task of comparing a learned posterior approximation $q(\theta \mid y)$ to the true posterior $p(\theta \mid y)$ can be framed as a two-sample test. Empirically, the Conformal C2ST outperforms classical discriminative tests across a wide range of benchmarks for this task. Our results establish the conformal C2ST as a practical, theoretically grounded diagnostic tool.
arXiv:2605.31572v1 Announce Type: new
Abstract: Reasoning is essential for autonomous driving (AD) in long-tail scenarios, where vehicles must apply commonsense knowledge, understand spatial relations, infer agent interactions, and make safe decisions. However, existing AD datasets and benchmarks mainly target perception, prediction, or planning, and provide limited supervision for reasoning over realistic long-tail driving scenes. We introduce nuReasoning, a large-scale real-world dataset and benchmark for reasoning-centric AD. Following the lineage of nuScenes and nuPlan, nuReasoning advances real-world AD datasets and benchmarks toward reasoning in long-tail driving scenarios. The dataset contains 20,000 clips, each 20 seconds long, collected across multiple cities, with synchronized multi-camera images, LiDAR data, HD maps, object annotations, and human-verified reasoning annotations spanning Spatial Reasoning, Decision Reasoning, and Counterfactual Reasoning. Unlike prior datasets that focus primarily on visual question answering, nuReasoning supports both reasoning evaluation and planning evaluation, enabling a direct study of how reasoning supervision affects driving performance. Experiments show that fine-tuning VLMs on nuReasoning substantially improves driving-specific question answering, while incorporating reasoning supervision into VLA training improves planning performance even when textual reasoning outputs are disabled at inference time. These results establish nuReasoning as a foundation for evaluating and improving robust, interpretable, reasoning-driven AD systems in realistic long-tail settings.
arXiv:2605.30592v1 Announce Type: new
Abstract: We study the problem of assigning a scalar score to a short trajectory window that reflects its position on an ordered continuum of predictability regimes, spanning structured deterministic dynamics to unstructured stochastic noise. Existing methods address deterministic-versus-stochastic discrimination within a single system and do not produce scores with a consistent numerical interpretation across systems. We formalize this as ordinal estimation over a five-level predictability ladder and identify a structural source of cross-system ambiguity: ranking supervision alone leaves the score coordinate unfixed up to a monotone reparameterization, which we term the gauge freedom of ordinal scoring. We propose the Gauge-Fixed Ordinal Network (GON), a temporal convolutional model trained with an anchor-and-variance objective that pins level-wise score means to shared target coordinates. GON operates on 2-jet features that expose local trajectory geometry, preserved by smooth flows and disrupted by stochastic surrogate procedures. On five held-out dynamical systems, initializing from a pretrained GON checkpoint consistently outperforms training from scratch across all window budgets, with adaptation depth reflecting geometric proximity to the training family. Zero-shot scores retain ordinal structure at the stochastic boundary, where surrogate procedures most strongly disrupt nonlinear geometry, and pretrained initialization consistently beats scratch across all window budgets. Pairwise discrimination and globally coherent ordinal scoring are distinct properties requiring a stable score coordinate for cross-system transfer, with direct implications for predictability assessment, model selection, and early-warning diagnostics across natural and engineered dynamical systems.
arXiv:2605.30659v1 Announce Type: new
Abstract: Materials used in polymer-based additive manufacturing processes, such as Digital Light Processing (DLP) and direct ink writing (DIW), typically exhibit non-Newtonian rheology. Carreau--Yasuda and power-law models describe basic shear-thinning and shear-thickening behavior well, but applying them to a new material requires choosing a functional form, deriving it, and re-implementing it inside the flow solver. We present a deployment workflow in which a neural network trained on experimental rheometry data serves as the viscosity closure inside a Cahn--Hilliard--Navier--Stokes (CHNS) finite element solver. Lipschitz regularization during training produces smooth viscosity predictions, and the trained network is exported in the Open Neural Network Exchange (ONNX) format and queried by the solver at runtime via the ONNX runtime, without solver modification or network reimplementation. The framework is built on a parallel octree-based adaptive mesh refinement infrastructure that concentrates resolution at the fluid interface. We validate the CHNS solver against benchmark shear-thinning bubble-rise cases from the literature, reproducing reported bubble shapes across varying power-law indices and Weber numbers. We characterized two silicone ink formulations, recorded their rise dynamics in perfluorodecalin on high-speed video, and used the resulting data to test the full workflow. Simulated rise velocities fall within the experimentally measured spread, and the simulated steady-state droplet shape agrees with the observed one. This work contributes to a growing body of literature on integrating neural constitutive closures into multiphysics simulations, and demonstrates a practical path for deploying experimentally trained rheological surrogates inside finite element solvers.
arXiv:2605.30460v1 Announce Type: new
Abstract: Accurate loss characterization is essential for the design of high-frequency power magnetic components. State-of-the-art resonant characterization methods are attractive for high accuracy and low sensitivity, especially at the MHz regime. However, they predominantly rely on manual tuning and computationally intensive Fast Fourier Transform (FFT) analysis to identify resonant conditions, causing both inefficiencies and inaccuracies. To ensure accuracy and expedite the process, this paper proposes a fully automated measurement architecture, the core innovation of which lies in the integration of digitally-controlled switched capacitor sequences and onboard signal processing circuits,enabling automated sweeping of both frequency and drive level for complete and rapid characterization with no human intervention. A design guideline for the switched capacitor sequence is presented and common commercial electromechanical power relays are characterized to enable sub-MHz measurements. Experimental results for several different magnetic materials demonstrate that the proposed system has great accuracy and is able to collect more than 1000 data points within 20 seconds, providing a very fast and robust solution for high-frequency magnetic characterization.
arXiv:2605.30777v1 Announce Type: new
Abstract: Autonomous coding agents built on large language models (LLMs) are rapidly being integrated into development workflows, yet their operational safety properties remain poorly understood beyond evaluations of explicitly malicious inputs. In practice, high-impact failures arise during benign, goal-directed use through environment breakage, fabricated success reports, etc. that current benchmarks do not capture. What categories of operational safety failures actually occur when coding agents are used for everyday development tasks and what is their impact? We present an incident-driven empirical study grounded in two complementary evidence streams. We screen 68,816 papers from 22 premier venues, curating 185 safety-relevant studies, and mine 16,586 GitHub issues from widely deployed LLM-powered coding tools, manually confirming 547 genuine safety failures. Applying systematic open coding over both corpora, we derive a multi-dimensional safety taxonomy of 33 operational risk types organized across seven dimensions, and annotate each incident with contributing factors, task context, severity, and downstream impact. Our findings show that coding-agent failures are often severe, with 326 of 547 incidents rated high or critical. The dominant risks are constraint violations, destructive operations, authorization bypasses, and deception, and over 65% of incidents arise in bug fixing and setup or configuration, patterns largely missing from prior literature. These results have direct implications for SE tool designers and benchmark developers: guardrails must go beyond adversarial-prompt defenses to enforce environmental constraints, failure transparency, and safe-halt behaviors.
arXiv:2602.07928v2 Announce Type: replace
Abstract: Flow-based generative models can be viewed through a physics lens: sampling transports a particle from noise to data by integrating a learned velocity field, and each sample corresponds to a trajectory with its own dynamical effort. Motivated by classical mechanics, we introduce Kinetic Path Energy (KPE), an action-like, per-sample diagnostic that measures the accumulated kinetic effort along an ordinary differential equation (ODE) trajectory. Empirically, KPE exhibits two robust correspondences: {i} higher KPE predicts stronger semantic fidelity; {ii} high-KPE trajectories land in sparse representation regions. We further provide theoretical guarantees linking trajectory energy to data sparsity. Paradoxically, this correlation is non-monotonic. At sufficiently high energy, generation can degenerate into memorization. Leveraging the closed-form formula of empirical flow matching, we show that extreme energies drive trajectories toward near-copies of training examples. This yields a Goldilocks principle and motivates Kinetic Trajectory Shaping (KTS), a training-free two-phase inference strategy that boosts early motion and enforces a late-time soft landing, reducing memorization and improving generation quality across benchmark tasks.
arXiv:2603.17145v2 Announce Type: replace
Abstract: Large language models (LLMs) are increasingly deployed as automated evaluators that assign numeric scores to model outputs, a paradigm known as LLM-as-a-Judge. However, standard Reinforcement Learning (RL) methods typically rely on binary rewards (e.g., 0-1 accuracy), thereby ignoring the ordinal structure inherent in regression tasks; for instance, they fail to recognize that predicting 4 is significantly better than predicting 1 when the ground truth is 5. Conversely, existing regression-aware approaches are often confined to Supervised Fine-Tuning (SFT), limiting their ability to explore optimal reasoning paths. To bridge this gap, we propose \textbf{REAL} (\underline{RE}gression-\underline{A}ware Reinforcement \underline{L}earning), a principled RL framework designed to optimize regression rewards, and also proven to be optimal for correlation metrics. A key technical challenge is that the regression objective is explicitly policy-dependent, thus invalidating standard policy gradient methods. To address this, we employ the generalized policy gradient estimator, which naturally decomposes optimization into two complementary components: (1) exploration over Chain-of-Thought (CoT) trajectory, and (2) regression-aware prediction refinement of the final score. Extensive experiments across model scales (8B to 32B) demonstrate that REAL consistently outperforms both regression-aware SFT baselines and standard RL methods, exhibiting significantly better generalization on out-of-domain benchmarks. On Qwen3-32B specifically, we achieve gains of +8.40 Pearson and +7.20 Spearman correlation over the SFT baseline, and +18.30/+11.20 over the base model. These findings highlight the critical value of integrating regression objectives into RL exploration for accurate LLM evaluation.
arXiv:2510.11683v3 Announce Type: replace
Abstract: A key challenge in applying reinforcement learning (RL) to diffusion large language models (dLLMs) is the intractability of their likelihood functions, which are essential for the RL objective, necessitating corresponding approximation during training. While existing methods approximate the log-likelihoods by their evidence lower bounds (ELBOs) via customized Monte Carlo (MC) sampling, they incur significant memory overhead due to the need to retain all MC samples for the gradient computation of non-linear terms in the RL objective, and thus restrict feasible sample sizes, leading to imprecise likelihood approximations and distorted RL objective. To address this, we propose \emph{Boundary-Guided Policy Optimization} (BGPO), a memory-efficient RL algorithm that maximizes a specially constructed lower bound of the ELBO-based objective. This lower bound is carefully designed to satisfy two key properties: (1) Linearity: it is a linear sum where each term depends only on a single MC sample, thereby enabling gradient accumulation across samples and ensuring constant memory usage; (2) Equivalence: Both the value and gradient of this lower bound are equal to those of the ELBO-based objective in on-policy training, making it also an effective approximation for the original RL objective. These properties allow BGPO to adopt a large MC sample size, improving likelihood approximations and RL objective estimation, which in turn leads to enhanced performance. Experiments show that BGPO significantly outperforms previous RL algorithms for dLLMs in math problem solving, code generation, and planning tasks. Our codes and models are available at \href{https://github.com/THU-KEG/BGPO}{https://github.com/THU-KEG/BGPO}.
arXiv:2602.11216v2 Announce Type: replace
Abstract: Molecular dynamics (MD) is a central computational tool in physics, chemistry, and biology, enabling quantitative prediction of experimental observables as expectations over high-dimensional molecular distributions such as Boltzmann distributions and transition densities. However, conventional MD is fundamentally limited by the high computational cost required to generate independent samples. Generative molecular dynamics (GenMD) has recently emerged as an alternative, learning surrogates of molecular distributions either from data or through interaction with energy models. While these methods enable efficient sampling, their transferability across molecular systems is often limited. In this work, we show that incorporating auxiliary sources of information can improve the data efficiency and generalization of transferable implicit transfer operators (TITO) for molecular dynamics. We find that coarse-grained TITO models are substantially more data-efficient than Boltzmann Emulators, and that incorporating protein language model (pLM) embeddings further improves out-of-distribution generalization. Our approach, PLaTITO, achieves state-of-the-art performance on equilibrium sampling benchmarks for out-of-distribution protein systems, including fast-folding proteins. We further study the impact of additional conditioning signals such as structural embeddings, temperature, and large-language-model-derived embeddings on model performance.
arXiv:2605.18024v2 Announce Type: replace
Abstract: Cooperation is central to multi-agent reinforcement learning (MARL), yet learned coordination can be fragile when external perturbations disrupt inter-agent interactions. Prior robust MARL methods have primarily considered value-oriented attacks, leaving a gap in robustness when interaction structures themselves are corrupted. In this paper, we propose an interaction-breaking adversarial learning (IBAL) framework that takes an information-theoretic view to construct attacks that impede coordination by perturbing agents' observations and actions, and trains agents to perform reliably under such disruptions. Empirically, our approach improves robustness over existing robust MARL baselines across diverse attack settings and yields stronger performance even under agent-missing scenarios. Our code is available at https://sunwoolee0504.github.io/IBAL.
arXiv:2602.01914v4 Announce Type: replace
Abstract: Token attribution methods provide intuitive explanations for language model outputs by identifying causally important input tokens. However, as modern LLMs increasingly rely on extended reasoning chains, existing schemes face two critical challenges: (1) efficiency bottleneck, where attributing a target span of M tokens within a context of length N requires O(M*N) operations, making long-context attribution prohibitively slow; and (2) faithfulness drop, where intermediate reasoning tokens absorb attribution mass, preventing importance from propagating back to the original input. To address these, we introduce FlashTrace, an efficient multi-token attribution method that employs span-wise aggregation to compute attribution over multi-token targets in a single pass, while maintaining faithfulness. Moreover, we design a recursive attribution mechanism that traces importance through intermediate reasoning chains back to source inputs. Extensive experiments on long-context retrieval (RULER) and multi-step reasoning (MATH, MorehopQA) tasks demonstrate that FlashTrace achieves over 130x speedup over existing baselines while maintaining superior faithfulness. We further analyze the dynamics of recursive attribution, showing that even a single recursive hop improves faithfulness by tracing importance through the reasoning chain.
arXiv:2605.31251v1 Announce Type: new
Abstract: Multimodal large language models (MLLMs) have shown strong potential as embodied agents, yet embodied geo-localization remains underexplored due to the lack of fine-grained evaluation. We introduce ERGeoBench, a diagnostic benchmark for vision-driven embodied geo-localization. ERGeoBench evaluates models under three progressive settings -- single-view, panorama-view, and embodied-view -- where agents may actively acquire observations through sequential changes in yaw, pitch, and zoom. The benchmark contains 2,207 globally distributed street-view panoramas and measures four complementary capabilities: foundational perception, spatial awareness, common sense reasoning, and geo-localization reasoning. Evaluations of leading proprietary and open-source MLLMs show that current models can infer high-level geographic semantics, but still struggle with fine-grained perceptual operations, metric localization, and spatial consistency across views. We further observe that geo-localization is strongly correlated with the other capability dimensions, suggesting that accurate localization depends on integrated perception, spatial reasoning, and commonsense inference rather than isolated visual recognition. Overall, ERGeoBench provides a unified framework for diagnosing and advancing human-like embodied geo-localization. Project Page: https://kaixuewen.github.io/ERGeoBench/
arXiv:2605.31294v1 Announce Type: new
Abstract: Recent advances in Audio-LLMs like GPT-4o have ushered in an era of conversational interaction with language models. Conversational avatars however, still seem robotic in facial expression and conversational flow, in part due to sequential stages of speech recognition, text generation, turn-based text response, speech synthesis, and audio driven facial animation. Based on our insight that audio-tokens produced by current Audio-LLMs carry sufficient information to reconstruct a plausible facial performance, we present TokTalk, a system that directly outputs expressive facial animation in real-time from streaming audio-tokens. We construct a novel audio-token to 3D facial motion dataset, on which TokTalk is trained using a Chunk-based Conditional Flow Matching model. A lightweight adaptation strategy allows our trained model to seamlessly connect to any token-based Audio-LLM at minimal computational overhead. Our chunk-based processing further enables parametric trade-off between latency and facial quality, shown through ablation studies. We further show that the real-time performance of TokTalk is comparable in latency to prior art solutions, and significantly favorable (via a perceptual study) in terms of quality, expressivity and control of the 3D facial performance. We showcase TokTalk's flexibility using a chatbot Avatar, a voice-driven user Avatar, and an animation Director's interface, as diverse audio-visual face applications.