arXiv:2607.05281v1 Announce Type: cross Abstract: Present-day quantum computing is cloud-based, where a user submits a circuit to a service provider's proprietary backend hardware. While providers may wish to hide implementation details, scheduling choices, or even which physical device was used, noisy finite-shot outputs can carry backend-specific fingerprints: information imprinted in the classical output distribution that can reveal the backend identity. So far, such fingerprints have mostly been studied from a benchmarking perspective, with limited attention to privacy considerations for users and providers. This work develops the first formal framework for backend identifiability and its privacy implications. We introduce a backend-identifiability game and use it to formalise routing anonymity as a security notion for quantum cloud services. We show that backend identifiability is a hypothesis-testing problem and prove that, under passive i.i.d. access to a single backend, routing anonymity decays exponentially at the Chernoff rate. We also establish a utility-anonymity trade-off, imposing fundamental limits on how much backend-specific information can be removed from classical outputs without degrading their usefulness. In addition, we observe that, for noisy quantum hardware, identifying fingerprints are inherently an intermediate-depth phenomenon, and establish a depth principle using Pauli-transfer-matrix tools. We complement the theory with experiments on Amazon Braket on AWS, using ion-trap and superconducting quantum processors. We observe 87-90% classification between superconducting backends and 96-100% classification across physical platforms, and find that identifiability can survive natural forms of post-processing. Overall, these results establish routing anonymity as a distinct security requirement for quantum cloud computing, and provide a framework for quantifying and controlling the utility-anonymity trade-off.
Science Journals
arXiv:2607.04718v1 Announce Type: new Abstract: Deep research agents decompose open-ended queries into subtasks, retrieve web evidence over multiple rounds, and synthesize long-form reports. This workflow creates a planning-layer poisoning surface: adversarial documents that enter the retrieval pool can steer follow-up questions and turn a local injection into report-level contamination. We present FORGE (Fabricated Orchestrated Reasoning chain for aGent Exploitation), a two-level attack that combines intra-document reasoning fabrication with inter-document chain coordination to hijack subtask planning. We further introduce the PRISM metric, which weights infected report claims by cognitive type, and Root Query Anchoring, a lightweight defense that ties recursive follow-up generation to the root query. Across 25 queries, Network FORGE reaches 26.4% PRISM with five injected documents and exhibits depth migration, in which recursive synthesis shifts poisoned content from overt framing into factual premises. On the 10-query defense subset, RQA (Root Query Anchoring) reduces PRISM from 38.5% to 18.3%.
arXiv:2601.15291v2 Announce Type: replace Abstract: Independent navigation is central to social participation and health for vulnerable populations. While historic cities such as Edinburgh often feature well-established public transport systems, urban accessibility challenges remain and are exacerbated by complex landscapes, especially for groups with multiple vulnerabilities such as the visually impaired elderly. With limited research examining how real-time data feeds and artificial intelligence in this context, we address this gap through a mixed-methods approach. Our spatio-temporal analyses make use of statistical and machine learning techniques to investigate network coverage, service patterns, and density profiles through live-recorded data. This is combined with a qualitative thematic analysis of semi-structured interviews with the target group, as well as links to spatial cognition theory. The results demonstrate the highly centralised nature of the city's transport system, the significance of memory-based navigation, and the lack of travel information in usable formats. We also find that participants already use navigation technology to varying degrees and express a willingness to adopt artificial intelligence. Our findings highlight the importance of dynamic tools to meaningfully improve independent travel, as well as limitations due to the recurring problem of specific accessibility data, for example for facilities, often not being collected and stored.
arXiv:2601.16510v4 Announce Type: replace Abstract: Solving massive-scale optimization problems requires scalable first-order methods with low per-iteration cost. This tutorial highlights a shift in optimization: using differentiable programming not only to execute algorithms but to learn how to design them. Modern frameworks such as PyTorch, TensorFlow, and JAX enable this paradigm through efficient automatic differentiation. Embedding first-order methods within these systems allows end-to-end training that improves convergence and solution quality. Guided by Fenchel-Rockafellar duality, the tutorial demonstrates how duality-informed iterative schemes such as the alternating direction method of multipliers, and the primal-dual hybrid gradient can be learned and adapted through representative case studies.
arXiv:2601.16806v4 Announce Type: replace Abstract: Insect neuroethology provides a compelling biological template for efficient autonomous navigation. We draw an analogy between the formal embodied AI visual point-goal navigation task and the ability of insects to discover, learn, and refine visually guided paths around obstacles between a discovered food location and their nest. We develop a novel integrative model of mushroom body and central complex, two insect brain structures, that have been implicated, respectively, in associative learning and path integration. We demonstrate the mushroom body learning triggered by collisions results in adaptive obstacle avoidance and consequently optimised paths to the goal, corroborating the hypothesis of recent behavioural work that an insect can learn continuously as they travel. The embodied insect-inspired model achieves success rates comparable to recent state-of-the-art models at many orders of magnitude less computational cost in the standardised Habitat point-goal navigation benchmark. Testing in a more realistic simulated environment validates its robustness to perturbations.
arXiv:2601.22912v2 Announce Type: replace Abstract: This paper studies integrated sensing and communication (ISAC) coordination for feedback control tasks under shared platform constraints. We consider a cyber-physical system in which a remote dynamical process (i.e., remote source) is regulated with the support of an ISAC-enabled base station that alternates between sensing the source state and communicating control-relevant information to the source, with the two operations semantically intertwined rather than serving separate targets and users. For a Gauss-Markov source with Bernoulli-distributed sensing and communication links and a finite-horizon linear-quadratic-Gaussian (LQG) cost, we derive the optimal ISAC and control policies. Under a Bellman-operator condition, we prove that the optimal ISAC policy at the base station follows an order-threshold structure in terms of the source and base-station estimation covariances, while the optimal control policy at the source follows a certainty-equivalent structure in terms of the source state estimate. We show that the threshold region, defined as the set of estimation covariance pairs for which communication is preferred over sensing, expands with increasing source uncertainty and contracts with increasing base-station uncertainty. Our numerical analysis validates the theoretical findings.
arXiv:2607.04881v1 Announce Type: new Abstract: Recently, omnitrees were introduced as a flexible space partitioning tree that improves upon the benefits of both octrees and k-d trees: Omnitrees' efficient encoding of anisotropic refinements holds particular interest for applications with anisotropic features and high dimensionality. These include, but are not limited to, computer graphics, databases, machine learning, and physics simulations. The present paper defines new operations on the omnitree encoding that extend its capabilities from the existing refinement to also include coarsening and therefore fully adaptive compression. It demonstrates natural integration of omnitrees with wavelets, which conserves moments of the stored function by design. For omnitrees, the wavelet coefficients can be interpreted as local refinement priorities, which can be used to guide the adaptation process. We derive algorithms for coarsening and downsplit that are guided by wavelet coefficients, and show their application to a large dataset of 3D shapes, as well as the continuous-valued density field of a cloud. The comparison to OpenVDB, a widely-used data structure for sparse volumetric data in computer graphics, enables a demonstration of the practical benefits of omnitrees even for moderately anisotropic three-dimensional data. Compared to OpenVDB, objects can be stored using up to 28x less space, and asymptotically show savings that exceed theoretical expectations. Using lossy compression, the cloud dataset can be compressed by $\approx5\times$ compared to OpenVDB, with negligible loss of visual quality. This demonstrates the potential of omnitrees for efficient storage and processing, and motivates further research into their applications in various domains.
arXiv:2607.04437v1 Announce Type: new Abstract: Using fully kinetic simulations that capture unprecedentedly large (from electron to ion) scales, we study magnetogenesis driven by continuous large-scale forcing until nonlinear dynamo saturation. We uncover a two-stage mechanism in collisionless ion-electron plasmas whose dynamics diverge dramatically from the pair-plasma case. In the first phase, electron pressure anisotropy triggers electron-Weibel modes, seeding small-scale magnetic fields. Then, a second growth phase emerges when the more massive ions develop their own strong anisotropy and drive ion-Weibel-type modes; concurrently, a Biermann-battery mechanism contributes to amplifying the magnetic field. This combined dynamics provides a tenfold amplification of the magnetic field in comparison to the pair-plasma case. Over long times, dynamo action continues until the system reaches a statistical steady state. This self-consistent kinetic mechanism provides a plausible explanation for robust magnetogenesis wherever an external forcing continuously stirs the plasma.
arXiv:2607.04637v1 Announce Type: new Abstract: Vision-Language-Action Models (VLAs), which leverage the advanced reasoning capabilities of Vision-Language Models (VLMs), show promising generalization in complex autonomous driving scenarios. Existing VLAs typically predict and optimize 3D trajectories from 2D images. While intuitive, this 2D-to-3D prediction is inherently entangled with camera parameters, leading to limited data scalability across heterogeneous driving datasets. Moreover, directly optimizing in 3D space induces severe convergence to trivial solutions, where VLAs rely on ego-status rather than visual scene understanding. To address these issues, we propose PixelPilot, a novel VLA featuring a decoupled planning and lifting paradigm. In the planning phase, PixelPilot reformulates scene understanding and trajectory prediction as sensor-agnostic 2D-to-2D tasks in the image plane, thereby facilitating scalable training across diverse datasets. The planned 2D trajectories are then deterministically lifted to 3D only during inference, ensuring the full exploitation of visual cues and generalization across different vehicles. To realize this paradigm, we propose a knowledge-instilled policy learning strategy that applies dense, intermediate rewards via Group Relative Policy Optimization (GRPO) to enforce a rigorous causal chain from visual perception to spatial planning. Extensive experiments demonstrate that PixelPilot achieves state-of-the-art performance in both open-loop and closed-loop settings, validating its superior scalability and visual reasoning capabilities.
arXiv:2602.04853v2 Announce Type: replace Abstract: Large language models often struggle to recognize their knowledge limits in closed-book question answering, leading to confident hallucinations. While decomposed prompting is typically used to improve accuracy, we investigate its impact on reliability. We evaluate three task-equivalent prompting regimes: Direct, Assistive, and Incremental, across different model scales and multi-hop QA benchmarks. We find that although accuracy gains from decomposition diminish in frontier models, disagreements between prompting regimes remain highly indicative of potential errors. Because factual knowledge is typically stable while hallucinations are stochastic, cross-regime agreement provides a precise signal of internal uncertainty. We leverage this signal to implement a training-free abstention policy that requires no retrieval or fine-tuning. Our results show that disagreement-based abstention outperforms standard uncertainty baselines as an error detector, improving both F1 and AUROC across settings. This demonstrates that decomposition-based prompting can serve as a practical diagnostic probe for model reliability in closed-book QA.
arXiv:2602.06371v2 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly deployed in multilingual contexts, yet their consistency across languages on politically sensitive topics remains understudied. This paper presents a systematic bilingual benchmark study examining how 17 LLMs respond to questions concerning the sovereignty of the Republic of China (Taiwan) when queried in Chinese versus English. We discover significant language bias -- the phenomenon where the same model produces substantively different political stances depending on the query language. Our findings reveal that 15 out of 17 tested models exhibit measurable language bias, with Chinese-origin models showing particularly severe issues including complete refusal to answer or explicit propagation of Chinese Communist Party (CCP) narratives. Notably, only GPT-4o Mini achieves a perfect 10/10 score in both languages. We propose novel metrics for quantifying language bias and consistency, including the Language Bias Score (LBS) and Quality-Adjusted Consistency (QAC). Our benchmark and evaluation framework are open-sourced to enable reproducibility and community extension.
arXiv:2602.06872v2 Announce Type: replace Abstract: We derive a residual-based $hp$-a posteriori error estimator for hybrid high-order (HHO) methods on simplicial meshes applied to the biharmonic problem posed on two- and three-dimensional polytopal Lipschitz domains. The a posteriori error estimator hinges on an error decomposition into conforming and nonconforming components. To bound the nonconforming error, we use a $C^1$-partition of unity constructed via Alfeld splittings, combined with local Helmholtz decompositions on vertex stars where the key contribution is to show that the stability constant only depends on the mesh shape-regularity. For the conforming error, we design two residual-based estimators, each associated with a specific interpolation operator. In the first setting, the upper bound on the conforming error involves only the stabilization term and the data oscillation, but hinges on an assumption that we verify numerically. In the second setting, the bound additionally incorporates bulk residuals, normal flux jumps, and tangential jumps. Numerical experiments confirm the theoretical findings on the error upper bound and also illustrate numerically that the proposed estimators lead to moderate effectivity indices.
arXiv:2602.06934v5 Announce Type: replace Abstract: Grassroots Logic Programs (GLP) is a concurrent logic programming language in which logic variables are partitioned into paired readers and writers. An assignment is produced at most once via a writer and consumed at most once via its paired reader, and may contain additional readers and/or writers. This enables the concise expression of rich multidirectional communication modalities. The language was introduced together with concurrent (cGLP) and multiagent (maGLP) operational semantics. Here, we derive from these (\ia)~dGLP, a deterministic counterpart of cGLP, and (\ib)~madGLP, a counterpart of maGLP in which deterministic agents communicate solely by asynchronous message passing, and prove them correct against their abstract counterparts. maGLP shared variable pairs spanning agents can be implemented as local variables paired by \emph{global links}, with correctness following from disjoint substitution commutativity (a consequence of GLP's single-occurrence invariant). We further prove that madGLP is grassroots. Both dGLP and madGLP serve as formal specifications for an AI-driven implementation discipline (math $\to$ informal spec $\to$ Dart) employed and described here: from dGLP, AI (Claude) developed a workstation-based GLP implementation in Dart, and from madGLP it is developing a smartphone-based multiagent one.
arXiv:2602.06941v3 Announce Type: replace Abstract: Large language models can recover mid-generation from task-misaligned activation steering, producing explicit verbal restarts (e.g., ``wait, that's not right'') and continuing on-topic even while the steering perturbation remains active. We term this Endogenous Steering Resistance (ESR). Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B exhibits explicit ESR at 3.8%, with smaller models from the Llama-3 and Gemma-2 families showing the explicit form less frequently. Two controls dissociate ESR into a detection event and a sustained-resistance component that conditioning on recent on-topic tokens does not fully explain. We identify 26 SAE latents through contrastive on-topic/off-topic search; zero-ablating them reduces the multi-attempt rate by 25%, with random-latent and held-out-prompt controls supporting specificity. ESR can also be deliberately enhanced through both meta-prompting and fine-tuning on synthetic self-correction examples. ESR has dual implications for safety: it could harden models against adversarial activation-space manipulation, but may equally interfere with beneficial steering-based interventions, since the model has no way to distinguish the two. Code is available at https://github.com/agencyenterprise/endogenous-steering-resistance.
arXiv:2607.04515v1 Announce Type: new Abstract: Efik, a tonal language spoken by about 3 million second language speakers and 1.5 million native speakers in Southeastern Nigeria, remains underrepresented in speech synthesis research. We present the first documented end-to-end text-to-speech study for Efik, introducing a curated single speaker corpus of 2,632 utterances totaling three hours and a comparative evaluation of four neural models (VITS, MMS-TTS, SpeechT5, and Orpheus-TTS) under low resource conditions. Native speakers evaluated the systems using MOS, Nat-MOS, and A-MOS. MMS-TTS achieved the highest MOS of 3.80 +/- 0.63 and produced more stable long form speech, though tonal errors persisted. Other models showed greater tonal and prosodic inconsistencies. These results provide a reproducible baseline and highlight the need for larger corpora and tone aware modeling for tonal African languages.
arXiv:2602.07789v2 Announce Type: replace Abstract: Rod-based structures are commonly used in practical applications in science and engineering. However, in many design, analysis, and manufacturing tasks, handling the rod-based structures in three dimensions directly is generally challenging. To simplify the tasks, it is usually more desirable to achieve a two-dimensional representation of the rod-based structures via some suitable geometric mappings. In this work, we develop a novel method for computing a low-distortion planar embedding of rod-based structures. Specifically, we identify geometrical constraints that aim to preserve key length and angle quantities of the 3D rod-based structures and prevent the occurrence of overlapping rods in the planar embedding. Experimental results with a variety of rod-based structures are presented to demonstrate the effectiveness of our approach. Moreover, our method can be naturally extended to the design and mapping of hybrid structures consisting of both rods and surface elements. Altogether, our approach paves a new way for the efficient design and fabrication of novel three-dimensional geometric structures for practical applications.
arXiv:2607.04570v1 Announce Type: new Abstract: Low Earth Orbit (LEO) satellite communications (SATCOM) has emerged as a key enabler of global connectivity for 6G networks. To overcome the significant path loss of space-to-ground links, high-gain directional beamforming (BF) is indispensable. As LEO systems evolve toward wider bandwidths to support data-intensive applications, however, they encounter a fundamental physical limitation known as the beam-squint effect, which induces frequency-dependent beam misalignment. Conventionally, the beam-squint effect has been treated as a critical performance impairment that must be mitigated. This article introduces a paradigm shift in wideband LEO satellite systems by redefining beam-squint as a valuable source of frequency-spatial diversity and presents the principles of rainbow BF. Rather than mitigating beam squint, rainbow BF deliberately exploits it to generate frequency-dependent beams, enabling different frequency components to illuminate distinct spatial directions using only a single or a small number of radio frequency chains. By supporting dynamic frequency-spatial beam allocation, rainbow BF offers enhanced flexibility and scalability for wideband LEO SATCOM. We further illustrate the benefits of rainbow BF through three representative LEO SATCOM applications: i) massive multiple access to overcome the latency and throughput bottlenecks of conventional beam hopping; ii) integrated sensing and communications for simultaneous target detection and data transmission; and iii) rapid satellite acquisition to reduce search overhead and improve link reliability. Finally, we discuss key implementation challenges and outline promising future research directions for rainbow BF in wideband LEO SATCOM.
arXiv:2204.13492v5 Announce Type: replace Abstract: We present StreamDEQ, a method that aims to infer frame-wise representations on videos with minimal per-frame computation. Conventional deep networks perform feature extraction from scratch at each frame in the absence of ad-hoc solutions. We instead aim to build streaming recognition models that can natively exploit temporal smoothness between consecutive video frames. We observe that the recently emerging implicit layer models provide a convenient foundation to construct such models, as they define representations as the fixed points of shallow networks, which need to be estimated using iterative methods. Our main insight is to distribute the inference iterations over the temporal axis by using the most recent representation as a starting point at each frame. This scheme effectively recycles the recent inference computations and greatly reduces the required processing time. Through extensive experimental analysis, we show that StreamDEQ is able to recover near-optimal representations within a few frames and maintain an up-to-date representation throughout the video duration. Our experiments on video semantic segmentation, video object detection, and human pose estimation in videos show that StreamDEQ achieves on-par accuracy with the baseline while providing 2-4x higher throughput.
arXiv:2603.13259v3 Announce Type: replace Abstract: When a decoder-only transformer is forced to process matched correct and incorrect single-token continuations of a factual query, the two pathways through hidden-state space diverge: displacement vectors from the query-only representation keep near-equal magnitude but rotate apart, with angular separation growing through mid-depth before late layers resolve an asymmetric outcome. A logit-lens preference in the incorrect run falls far below the equal-probability prior (roughly 11.5x more mass on the incorrect token than the correct one). We read this pattern, rotational divergence then late-layer asymmetric commitment, as the geometric signature of the model externally appearing to reject a wrong continuation, while staying explicit that it is observational, not causal: the incorrect run could equally reflect the model conforming to the token it is forced to carry, which only a random-token control can settle. It holds across six decoder-only transformers spanning four architecture families (Llama, Mistral, Gemma, StableLM) from 1B to 13B parameters; a seventh (Qwen2 1.5B) is flat under our protocol, plausibly a tokenizer artefact, leaving an emergence threshold open. Linear probes recover the distinction at intermediate depth, and cross-domain transfer is structurally asymmetric, a financial-medical corridor transferring far better than transport pairs. Where single-layer activation patching is cleanly interpretable (LLaMA-2 13B, Mistral 7B) it yields no layer band of consistent recovery; a third model (StableLM-2 1.6B) recovers uniformly above the ceiling, which we diagnose as a code-path artefact and exclude. Under this scoped null the late-layer asymmetry is not localized to a single component, fitting a distributed-by-trajectory account rather than single-layer localized recall. We document this with forced-completion probing across seven models, three domains, and 300 queries.
arXiv:2607.04783v1 Announce Type: new Abstract: Cellular data billing is a core operational mechanism for mobile Internet service providers (ISPs), and a policy gap that excludes a specific protocol from usage accounting can lead to a practical security threat. Some cellular ISPs treat ICMP echo traffic as control traffic rather than user data and exclude it from billing. At the same time, Android allows ordinary applications to create ICMP echo sockets without root privileges because of an unsafe default configuration, and the combination of these two conditions forms a vulnerability that can bypass data billing. Existing billing-bypass attacks either require root privileges to create raw sockets and modify routing tables, or do not provide an end-to-end implementation that works in a non-rooted environment, which limits the threat to a small group of experts. This paper proposes Ghost Traffic, an end-to-end system that uses Android's VpnService to encapsulate all application traffic into ICMP echo payloads without root privileges and route it through an external proxy server. The proposed system targets both public IPv4 environments and IPv6-only LTE environments through two variants: IPv4 ICMP tunneling and IPv4-over-IPv6 ICMP tunneling. We evaluated its applicability in seven ISP environments in South Korea, Japan, and the United States, and observed end-to-end tunneling in six of them. We observed that billing bypass occurred in multiple environments and quantitatively showed this effect by measuring that Quality of Service (QoS) throttling was not applied even after the data cap was exhausted. Finally, we propose layered countermeasures across the device, platform, and network levels, performed responsible disclosure, and show that the operational practice of not billing ICMP traffic can lead to practical billing bypass.
arXiv:2607.04786v1 Announce Type: new Abstract: Test maintenance is a critical, yet costly, activity - particularly as codebases rapidly evolve. To assist, we present MAST, a multi-agent framework that predicts which test cases require maintenance following changes to the production code. This identification task is necessary as a precondition to any subsequent maintenance activities, but remains challenging due to the complex relationships between production and test code. MAST advances the state-of-the-art by integrating multiple analyses -- including static, lexical, and semantic analyses - through an intelligent fusion and post-check procedure and by focusing on a realistic use and evaluation setting - i.e., standardized input formats, repository-level analyses, and the ability to infer relations between test and production artifacts rather than assuming a pre-existing mapping. We evaluated MAST on 21 industrial Java repositories from Ericsson AB, considering situations where test maintenance both was and was not required in the ground truth. MAST yielded superior precision to a state-of-the-art baseline - resulting in a higher accuracy, F1, and F2 score - with only some loss in recall. Our ablation study demonstrates the value of each analysis in producing the final recommendations. MAST illustrates the potential of multi-agent systems that can fuse multiple information sources when performing software testing tasks.
arXiv:2607.04789v1 Announce Type: new Abstract: Classical partially synchronous state machine replication, as in PBFT, tolerates f Byzantine replicas among n at least 3f+1 using three communication steps per request. Recent protocols such as Minimmit achieve two-message-delay decisions under stronger size assumptions, notably n at least 5f+1 when any silent replica must be counted as a potential equivocator. Hydrangea and Kudzu treat mixed Byzantine and crash faults, focusing on providing a fast-path under optimistic conditions while maintaining a fall-back commitment path similar to PBFT. In this paper, we also consider a mixed model, but focus on studying the fault tolerance of the 2-message-delay commit. For this, we prove a tight bound of n at least 5f+3c+1. Extending this result, we also show that there exists a more resilient commit path that allows an extra f_abc < n-3f-2c alive-but-corrupt faults at 4-message-delays. Core liveness is claimed in executions with at most f equivocators; if this regime is violated (e.g., AbC-induced forks), the protocol enters synchronous recovery, where only the resilient-path safety guarantee is preserved. As a result, for f=16, c=6, and n=99, we obtain a commit path that tolerates 22% of replicas failing for liveness, 16% equivocating for 1-RTT safety, and 54% equivocating for 2-RTT safety.
arXiv:2607.04800v1 Announce Type: new Abstract: Neural networks are thought to represent concepts as directions in their activation space, and superposition lets them encode more concepts than they have dimensions. It is natural to ask whether they can also compute more functions than they have neurons, i.e., perform computation in superposition. In this regime many functions of sparse inputs are evaluated by a layer with fewer neurons than there are functions to compute. Representation in superposition is by now fairly well understood, but computation in superposition is not, and there are few toy models of it arising through training rather than being hand designed. As a toy model of computation in superposition we study the compressed-computation setup: a single-hidden-layer ReLU network with 50 neurons that must compute the ReLU of each of 100 sparse input features. We show that training it under an $L^4$ loss (the mean fourth power of the error), rather than the usual $L^2$, elicits a solution that appears to compute all features in superposition. We then reverse-engineer this solution. We find that the network assigns each feature a sparse binary codeword over neurons and decodes it with a pseudoinverse of the encoder. Given these codewords, a description with only three scalars recovers most of the network's performance, and we validate it by building equivalent networks from hand-designed codes.
arXiv:2607.04854v1 Announce Type: new Abstract: Despite their strong reasoning capabilities and extensive world knowledge, Large Language Models (LLMs) frequently generate plans that violate task constraints, undermining their reliability in real-world applications. This deficiency arises from a lack of systematic mechanisms to incorporate constraint information during the generation process. While existing approaches attempt to mitigate this by relying on external tools or task decomposition, they fail to enhance the model's intrinsic constraint awareness. To address this, we propose Constraint-Aware Reinforcement Learning (CARL), a novel RL framework designed to strengthen LLMs' intrinsic focus on constraints. CARL introduces a constraint-aware reward by comparing the model's output distributions under constrained and unconstrained inputs, encouraging constraint focus and penalizing neglect. Compatible with various RL frameworks and requiring no external solvers or top models, CARL enables scalable, end-to-end constraint-aware planning. Extensive experiments on BlocksWorld, TravelPlanner, and T-Eval demonstrate that CARL significantly outperforms standard Reinforcement Fine-Tuning (RFT) baselines and state-of-the-art reasoning models, exhibiting a markedly increased focus on constraints.
arXiv:2607.04851v1 Announce Type: new Abstract: Generalizing sleep staging models to unseen datasets is challenging, and typical domain generalization (DG) methods often rely on multiple source domains or domain labels that are rarely available in practice. We tackle the stricter and more practical setting of single-source domain generalization: training on a single labeled source dataset, without domain labels or access to target data. We present SleepBand, a physiology-guided framework that embeds oscillatory priors via a learnable Morlet filter bank and a structured integration-and-recalibration pipeline. This anchors representations to domain-invariant sleep rhythms (e.g., slow waves, spindles), reducing reliance on dataset-specific artefacts. On five public datasets, SleepBand achieves state-of-the-art SDG performance and remains competitive under leave-one-domain-out (multi-source) DG. Analyses show that the learned filters align with canonical neurophysiology and that robustness stems from focusing on narrowband, physiologically meaningful cues. Our results suggest that principled, physiology-aware inductive biases are a promising path for robust single-domain sleep staging. Code is available at https://github.com/lzcn/sleep-band