arXiv:2606.22511v2 Announce Type: replace Abstract: In open-ended generation, LLMs frequently fall into the "likelihood trap", marked by repetitive degeneration and vocabulary dullness, creating a discrepancy between machine-generated and human-written text. While post-hoc tail truncation (e.g., Top-$p$, Min-$p$) avoids sampling from the unreliable tail, it can over-sample from the uncalibrated head and misalign generation with human lexical preferences; fixed scalar repetition penalties likewise ignore variation in logit scale across inference steps, potentially disrupting semantic coherence. To address both limitations, we propose Variance-Calibrated Modulation (VCM), a training-free pre-decoding intervention that reshapes the probability distribution before truncation through two dynamic mechanisms: (1) Contextual Searchlight via PMI, which suppresses global stopwords while elevating context-evoked tokens, and (2) Adaptive Self-Debiasing, which uses real-time logit standard deviation for scale-invariant penalization. Across open-ended generation, factual QA, and mathematical reasoning, VCM consistently mitigates the likelihood trap. With negligible computational overhead, VCM integrates with existing decoding strategies, improving diversity, coherence, and, particularly at higher decoding temperatures, reasoning accuracy.
Science Journals
arXiv:2607.05475v1 Announce Type: new Abstract: Deploying Large Language Models (LLMs) on mobile devices enhances privacy and reduces latency, but is severely bottlenecked by hardware inefficiency. We present the first comprehensive, cross-layer measurement study of mobile LLM inference, uniquely spanning five mainstream frameworks (e.g., llama.cpp, GENIE) and three hardware backends (CPU, GPU, NPU). To enable this analysis, we develop PowerBench, a fine-grained profiling tool that provides the first backend-specific energy attribution, moving beyond traditional device-level measurements. Our study yields three critical insights: (1) Framework-induced performance gaps are substantially amplified on NPUs, reaching up to 10x using custom operators due to divergent offloading and quantization strategies. (2) We identify a distinct phase split where NPUs excel at compute-bound prefilling, while CPUs outperform all other backends in memory-bound decoding. This is driven by the NPU's preference for large, fixed-shape workloads, which conflicts with the small-kernel, dynamic nature of decoding. (3) Backend-specific profiling uncovers substantial scheduling headroom missed by prior work. Suboptimal thread configurations, uncoordinated NPU sleep latencies, and CPU polling intervals result in up to 40% energy waste. Leveraging these findings, we present an energy-oriented best-practice configuration for mobile LLM inference. We estimate that this configuration could reduce energy consumption by up to 54.8% on the NPU backend across three datasets.
arXiv:2607.06466v1 Announce Type: new Abstract: This paper addresses the remote verification of the authenticity of Optically Variable Devices (commonly known as holograms) on identity documents. Typically placed over the cardholder's photo, these devices provide strong and easily verifiable security for human inspection but pose challenges for automated verification. Existing approaches easily cover static frauds (e.g. paper photocopy) and can be evaluated for such, but their capacity to detect real, dynamic fraud cases (e.g. handcrafted hologram) has not been evaluated to date because of the lack of public datasets. Furthermore, they are usually trained to detect known attack types, and few of them can generalize to new, unseen attacks. This work features three contributions to address these limitations: 1) a new public dataset, MIDV-DynAttack, which extends the existing MIDV-Holo dataset with realistic, static and dynamic attacks against identity document specimens, tripling the number of attack samples compared to the original dataset, 2) a novel verification method which can assess the authenticity of a specific hologram thanks to the analysis of its dynamic behavior and appearance, can be trained without dynamic attack samples, and exhibits new state-of-the-art performance, 3) a benchmark of existing approaches which follows a clear evaluation protocol and emphasizes the inability of other approaches to deal with dynamic attacks, as well as new challenging attacks to deal with. Code and dataset are publicly available at https://github.com/EPITAResearchLab/pouliquen.25.icdar.
arXiv:2607.05479v1 Announce Type: new Abstract: Generative AI (GenAI) systems store and process client data in three distinct ways: in the model's parameters through training and memorisation, in the context window during a live session, and in knowledge databases for retrieval-augmented generation (RAG). Each mode creates different and often counter-intuitive risks to confidentiality and legal professional privilege, and each calls for specific governance responses. Drawing on the first English and American decisions to address privilege and generative AI, UK and Munir v Secretary of State for the Home Department and United States v Heppner, on the orthodox privilege authorities against which those decisions must be read, and on recent computer science research, we explain the three modes of data storage and processing in terms accessible to practitioners and analyse the legal consequences of each. We then situate the analysis within the regulatory framework governing solicitors in England and Wales and within the ordinary principles of professional negligence, arguing that the standard of effective information governance (and with it the benchmark against which negligence and misconduct will be measured) is changing. Although we write primarily for SRA-regulated practitioners, our data-governance analysis is framed to extend to any jurisdiction in which the protection of privilege or professional secrecy depends on demonstrable confidentiality. The ultimate aim of this article is to help legal services professionals understand salient data leakage risks in GenAI systems and thereby facilitate a more responsible deployment of GenAI on client data and other sensitive material.
arXiv:2607.05482v1 Announce Type: new Abstract: Large language models (LLMs) can plan behavior for embodied agents from natural language, but treating the LLM as a request/response oracle on the critical path is fundamentally at odds with real-time control and concurrent goals. We argue for an operating-system-style runtime for embodied agents, and instantiate this idea in an early prototype, TypeGo. TypeGo structures LLM-based planning as asynchronous loops at multiple timescales that overlap with execution, and manages the agent's physical body like an OS manages hardware: the Skill Kernel arbitrates typed physical subsystems among concurrent per-task processes, a scheduler preempts them and resumes or replaces each by source, and speculative skill streaming hides LLM latency behind ongoing motion, while a fast first-action path yields visible feedback within a second. Users program behavior through natural language prescriptions that TypeGo dispatches to the LLM-based planners or compiles into low-latency interrupt handlers. Our prototype of Kalos, a Unitree Go2 quadruped, provides preliminary evidence for the design: in our current task suite, it cuts per-step delay by 50% over step-by-step planning and time-to-first-action by 73% over monolithic planning, while admitting concurrent tasks at low scheduling overhead.
arXiv:2606.22807v2 Announce Type: replace Abstract: As retrieval systems scale, high-quality reranking becomes increasingly important. However, most existing rerankers, whether encoder-based or decoder-based, jointly encode the query and passage, tightly coupling their computation and limiting deployment efficiency as well as flexibility. We present KaLM-Reranker-V1, a fast but not late-interaction (FBNL) reranker that decouples query and passage computation while retaining expressive relevance modeling. Built on an encoder-decoder architecture, KaLM-Reranker-V1 uses the encoder to pre-encode passages with Matryoshka embedding pooling, while the decoder models the system instruction, user instruction, and query intent; cross-attention then captures relevance between the query context and passage representations. This design makes KaLM-Reranker-V1 efficient through decoupled passage encoding, yet not late interaction, by preserving rich relevance modeling through cross-attention. We instantiate KaLM-Reranker-V1 in three sizes, Nano, Small, and Large, with 0.27B, 1B, and 4B activated parameters, respectively. Extensive experiments on BEIR, MIRACL, and LMEB demonstrate that KaLM-Reranker-V1 achieves strong reranking performance with superior efficiency. On BEIR, KaLM-Reranker-V1 achieves state-of-the-art performance, on par with strong industrial models such as the Qwen3-Reranker series; on MIRACL, despite not being extensively trained on multilingual data, KaLM-Reranker-V1 still shows excellent reranking performance. Moreover, on LMEB, reranking models demonstrate a clear advantage, with even the 0.27B Nano model remaining competitive with 7-12B embedding models.
arXiv:2607.05783v1 Announce Type: new Abstract: Environmental illusions (eg., shadows, reflections, and tire marks) are naturally existing yet overlooked phenomena in real-world driving environments. They can disturb visual perception, leading to misinterpretation of the scene and posing serious safety risks to autonomous driving (AD) systems. However, existing researches largely overlook these phenomena, leaving a critical gap. To address this issue, we study AD robustness through the lane perception perspective, a fundamental task supporting core functions like cruise control and lane centering. We focus on two representative models: conventional lane detection (LD) and vision-language model-based systems (ADVLMs). In this work, we introduce the first benchmark, LanEvil++, for evaluating the robustness of lane perception under environmental illusions. LanEvil++ encompasses 14 types of illusions and leverages the CARLA simulator to generate 94 high-fidelity, fully controllable 3D scenes, yielding a dataset of 90,292 annotated images, 1,596 video clips, and 41,855 visual question answering pairs. Extensive evaluations demonstrate that environmental illusions substantially degrade the performance of state-of-the-art LD methods. On average, LD models experience a 5.27% drop in Accuracy and a 10.49% decline in F1-score, while ADVLMs show a 2.03% reduction in GPT-score and a 0.75% drop in Language-score. Among all illusions, shadows emerge as the most disruptive factor, reducing accuracy by up to 7.20%. Furthermore, closed-loop simulations reveal that these illusions can lead to incorrect driving decisions. Complementary real-world case studies highlight safety-critical failures in actual traffic scenes. To enhance robustness, we propose the Multimodal Illusion Defense Approach (MIDA). MIDA achieves substantial gains under challenging conditions, boosting robustness by 4.23% on LD models and 3.82% on ADVLMs.
arXiv:2607.05533v1 Announce Type: new Abstract: Computational pathology foundation models (PFMs) have advanced whole-slide image analysis. However, their size and inference cost hinder local deployment in pathology departments. We propose MuCoDi, a pretraining framework that distills frozen tile embeddings from multiple PFMs into compact edge-oriented encoders. Instead of regressing individual teacher features, MuCoDi trains lightweight MobileOne and RepViT students with a contrastive distillation objective adapted from MoCo v3, where cached Virchow2, UNI2, and H-Optimus-1 embeddings replace momentum-encoder keys. We pretrain students on 14.3M TCGA tiles from only 11.8K WSIs and evaluate frozen encoders on 23 clinically curated downstream classification tasks. RepViT-based MuCoEdge students retain near-teacher performance while reducing model size by orders of magnitude: MuCoEdge-R2.3 and MuCoEdge-R1.5 reach 71.0% external AUROC, within 0.8 percentage points of the best teacher (Virchow2, 71.8%), while MuCoEdge-R2.3 obtains the best external F1 and the second-best AUPRC (51.8% and 53.3%). MuCoEdge-R1.0 reaches 70.9% AUROC with only 6.4M parameters and 1.12 GFLOPs. On a Raspberry Pi 5, sub-million-parameter MobileOne students achieve up to 605-fold single-tile speedup over Virchow2 while retaining 66.5% to 66.9% external AUROC, demonstrating that PFM-quality pathology representations can be moved toward practical edge deployment. Code is available at https://anonymous.4open.science/r/mucodi-6243.
arXiv:2607.05541v1 Announce Type: new Abstract: Reinforcement Learning is commonly used to train large language models using environmental feedback. In applied settings, the environment usually provides sparse or delayed feedback. This makes it difficult for the model to pinpoint which actions in its reasoning led to success or failure. So, learning effectively from these signals is hard because the model must determine how each failure should inform meaningful behavioral corrections in subsequent iterations. We introduce a training framework, Self-Review Reinforcement Learning, that embeds an explicit self-review step into each RL episode. When a first-pass response fails, the model generates a self-review to identify what went wrong, which conditions an improved second attempt. Unlike inference-time reflection approaches, such as Reflexion, the framework optimizes self-review with policy gradients and internalizes improvements into the base policy via selective distillation, ensuring they persist across future episodes. A cross-episode memory keeps successful self-reviews for reuse when encountering similar tasks in future episodes during training. We evaluate SRRL against a standard RLVR baseline using the GRPO optimizer across two language models, Qwen 3-4B and OLMo-3- 7B, on GSM8K benchmark. SRRL consistently outperforms the RLVR in final reward performance and achieves greater learning efficiency by successfully transforming feedback into behavioral improvement.
arXiv:2606.24234v2 Announce Type: replace Abstract: Autonomous robotic inspection in maritime environments presents unique challenges for Visual Place Recognition (VPR) due to cross-scene perceptual shifts. Robots navigating ship-borne environments must transition between visually distinct domains: open decks with sparse textures and severe illumination changes, and enclosed cabins with repetitive structures and high visual ambiguity. Existing VPR methods, designed primarily for urban or indoor scenes, fail to generalize reliably across these starkly different scenarios. To address this, we propose ProteusVPR, a two-stage retrieval-refinement framework. The first stage employs any standard VPR model for initial image retrieval. The second stage introduces a geometric-visual estimation network that fuses the retrieved image with two temporally preceding frames, incorporating geometric descriptors, a local affine coordinate system, and camera azimuth encoding to achieve precise localization. To support this task, we introduce the XHZ dataset, an 8K-panoramic ship-borne dataset collected from an operational vessel, featuring multi-floor cabin structures, deck transition zones, and strict query-database separation for rigorous evaluation. Extensive experiments on the XHZ dataset demonstrate that ProteusVPR consistently improves the localization accuracy across multiple VPR backbones, reducing mean localization error by over 60\% on average and that ProteusVPR offers an effective and robust solution for precise visual localization in challenging, cross-scene maritime environments.
arXiv:2606.27106v2 Announce Type: replace Abstract: We present a novel approach for applying Large Language Models (LLMs) to threat assessment in the context of foreign peacekeeping missions. Building on the PINPOINT project and its use case, the EU Monitoring Mission in Georgia, we combine an interdisciplinary risk-model with OSINT-based media collection and LLM-supported threat extraction. The proposed workflow maps media contents to mission-relevant threats, extracts structured information and applies several additional LLM-based processing steps to improve relevance and grounding. An evaluation of threats extracted from media documents shows high agreement between automatically generated results and human judgment for core aspects such as threat and mission relevance. These results indicate that LLMs provide a promising approach to support analysts in the context of peacekeeping missions.
arXiv:2606.28988v2 Announce Type: replace Abstract: Machine learning for underwater acoustics is constrained by the scarcity of publicly available labeled datasets. In contrast to air-acoustic domains, where large benchmarks enable rapid model development, underwater datasets are typically small and limited in acoustic diversity, restricting robust model training and cross-domain generalization. To help address this gap, we introduce a curated underwater audio dataset derived from an open-source maritime sound archive. The dataset contains over one thousand labeled audio segments across eight biologically and mechanically relevant acoustic classes, providing an additional resource for training models in data-limited underwater environments. Additionally, we establish a lightweight Convolutional Neural Network (CNN) baseline and propose a margin-enhanced loss with feature alignment to mitigate class confusion arising from data imbalance, acoustic similarity, and cross-domain mismatch. While the baseline achieves 96.35% in-domain accuracy, evaluation on ShipsEar reveals substantial domain shift; the proposed feature alignment improve zero-shot ship detection by 42.60%, demonstrating stronger robustness under distribution mismatch. We further release a transparent curation pipeline and reproducible benchmark to support future research on imbalance mitigation, domain adaptation, and data-efficient underwater acoustic classification.
FADE: Mitigating Hallucinations by Reducing Language-Prior Dominance in Large Vision-Language Models
arXiv:2606.29431v4 Announce Type: replace Abstract: Despite the impressive capabilities of Large Vision-Language Models (LVLMs), they remain susceptible to hallucination, generating content inconsistent with the input image. Recent studies attribute this to the dominance of language priors over visual inputs and employ contrastive decoding methods to mitigate this dominance, but the mechanistic origin remains unexplored. We investigate the information flow through each transformer layer and find that attention modules consistently aggregate visual evidence, while FFN modules at critical layers act as the source of language priors. These priors can override visual evidence, causing correct predictions in intermediate layers to drift toward incorrect outputs. Based on this insight, we propose FADE (FFN Attenuation for DEcoding), a training-free method that attenuates FFN outputs to reduce language-prior dominance. Evaluations on POPE, CHAIR, and MME benchmarks across LLaVA-1.5, mPLUG-Owl2, and InstructBLIP show that FADE effectively mitigates hallucinations while preserving inference efficiency.
arXiv:2606.30243v2 Announce Type: replace Abstract: This paper presents KYON, a hybrid wheel-legged quadruped robot equipped with a bimanual upper body for loco-manipulation tasks. The platform features a semi-modular design with a reconfigurable lower legs, enabling both wheeled and legged locomotion depending on the environment. A design approach that places actuators in the base and uses transmission mechanisms reduces distal inertia, improving agility and dynamic performance. The robot integrates a whole-body control framework together with a reinforcement learning based policy to handle nonlinear dynamics and enhance robustness to disturbances for the execution of locomotion and manipulation tasks, independently. Experimental results demonstrate effective dynamic locomotion and bimanual manipulation, validating the platform's capability to operate in complex and unstructured scenarios.
arXiv:2607.06091v1 Announce Type: new Abstract: The advection-diffusion equation is a fundamental model of transport phenomena in which mass conservation is an essential physical constraint. While classical schemes such as Crank-Nicolson preserve this property by construction, Physics-Informed Neural Networks (PINNs) enforce only the local residual of the governing PDE and are therefore not guaranteed to conserve global quantities such as mass over long integration horizons. In this work, we examine the extent of this limitation for the periodic one-dimensional advection-diffusion equation and evaluate a Mass-Penalty PINN that augments the standard PINN loss with a soft mass-conservation constraint. We compare the performance of Vanilla PINN, Mass-Penalty PINN, and the Crank-Nicolson scheme across a range of Peclet numbers spanning diffusion-dominated to advection-dominated regimes, and over two simulation horizons representing short-term and long-term dynamics. The results show that, for short-term simulations, the Mass-Penalty PINN does not always provide a consistent improvement in accuracy. However, for long-term simulations, the Mass-Penalty PINN reduces the relative L2 error and mass conservation error by factors of approximately 9-67 and 15-215, respectively, compared with the Vanilla PINN, across the tested Peclet numbers. Further analysis reveals that the accuracy degradation observed in Vanilla PINN is predominantly caused by the accumulation of mass drift over time. These results demonstrate that incorporating a soft mass-conservation constraint substantially improves the long-term reliability of PINN for conservative transport problems, particularly in mitigating mass drift over extended simulation horizons.
arXiv:2607.06496v1 Announce Type: new Abstract: This paper studies the constrained-capacity for precoded faster-than-Nyquist (FTN) signaling with finite-alphabet inputs. Despite the promise of accelerated transmission, the fundamental rate limit of precoded FTN signaling under practical finite-alphabet constraints remains unclear. By introducing cyclic prefix (CP) and cyclic suffix (CS), the FTN channel is decomposed into a set of parallel eigenchannels by the discrete Fourier transform (DFT) matrix, based on which the constrained capacity is derived. The results demonstrate that time acceleration can improve spectral efficiency over Nyquist signaling even when a fixed modulation order is employed. Moreover, in the low and moderate signal-to-noise ratio (SNR) regimes, a smaller constellation combined with stronger time acceleration can outperform a larger constellation with weaker acceleration. Next, the asymptotic behavior of the constrained capacity is analyzed as the acceleration factor tends to zero under both fixed transmit-SNR and fixed receive-SNR definitions. It is shown that the constrained capacity for DFT-precoded FTN is fundamentally limited by the constellation size. In addition, the constrained capacity under channel mismatch is studied and a mismatched achievable information rate (AIR) formulation is developed to show the effects of practical constraints on the performance degradation. Finally, adaptive bit loading across eigenchannels is investigated to exploit the higher-quality eigenchannels.
arXiv:2607.05726v1 Announce Type: new Abstract: Association unlearning aims to disable learned label-attribute shortcuts while preserving task performance. Existing evaluations mainly measure output-level robustness or probe whether shortcut attributes remain readable in frozen features, but neither test determines whether a retained association remains functionally usable by the original classifier. We propose the Association Restoration Test (ART), a post-hoc diagnostic for functional shortcut restorability. ART estimates class-conditional association directions, amplifies residual components, and evaluates the modified features with the original classifier head. Across Waterbirds, CelebA, SpuCoDogs, and an ISIC timestamp-artifact extension, we show that output metrics, representation probes, and ART characterize distinct aspects of shortcut mitigation. These findings motivate restoration-aware evaluation for unlearning and shortcut-mitigation methods that target learned associations rather than individual classes or concepts.
arXiv:2606.30340v2 Announce Type: replace Abstract: We formulate a Bayesian framework for reconstructing doping profiles in pn-junction semiconductor devices from boundary flux measurements. The unknown doping field is modeled as a piecewise-constant function characterized by an unknown interface and two plateau concentrations, leading to a nonlinear ill-posed inverse problem governed by a Poisson-Boltzmann-type equation. To represent this structure while enabling efficient gradient-based inference, we introduce a pushforward prior constructed by mapping a latent Gaussian field with Mat\'ern-type covariance through a sigmoid transformation. The latent field is parameterized by a truncated Karhunen-Lo\`eve expansion, while the two piecewise-constant levels are represented by scalar plateau parameters. The prior yields differentiable approximations of piecewise-constant fields with controllable interface sharpness. We establish well-posedness of the Bayesian formulation by proving Lipschitz continuity of the forward map and Hellinger stability of the posterior. We then sample the posterior using the No-U-Turn Sampler (NUTS) with gradients computed by the adjoint method. Numerical experiments show that the combination of the proposed prior and NUTS provides more efficient posterior exploration than the dimension-robust preconditioned Crank-Nicolson (pCN) sampler, yielding one to two orders of magnitude larger effective sample sizes. In the known-plateau setting, the method reconstructs both planar and curved interfaces and provides spatially resolved uncertainty quantification (UQ). When the interface geometry and plateau concentrations are inferred jointly, posterior correlations reveal structural non-identifiability. These results demonstrate the effectiveness of combining pushforward priors with adjoint-gradient-based sampling for reliable UQ in nonlinear partial differential equation-constrained inverse problems with sharp interfaces.
arXiv:2606.30530v5 Announce Type: replace Abstract: The continuous integral R2 indicator is a Pareto-compliant refinement of the classical finite-weight-vector R2 indicator, used in performance assessment, bounded archiving for a-posteriori multi-objective optimization, and skyline selection in databases. This work introduces a bidirectional perspective mapping between continuous integral R2 computation and integration over unions of anchored axis-aligned boxes. After translating the ideal point of a minimization problem to the origin, approximation points become strictly positive loss vectors, and the subgraph of the lower weighted Tchebycheff envelope over the weight simplex maps to the complement of an anchored-box union in reciprocal objective space. The Jacobian gives an absolute R2 formula as a weighted complement volume with density $(x_1+\cdots+x_N)^{-(N+1)}$, while differences of R2 values become finite weighted hypervolume differences. Hence, hypervolume algorithms that emit box decompositions can be reused by replacing ordinary box volumes with closed-form weighted box integrals. For $N$ objectives, this gives an output-sensitive overhead $O(2^N M)$ for an $M$-box decomposition, or $O(M)$ for fixed $N$. Using existing box-decomposition approaches, the integral R2 can be computed in $O(n \log n)$ for $N=2,3$, in $O(n^2)$ for $N=4$, and in $O\left(n^{\lfloor (N-1)/2\rfloor+1}\right)$ for $N\geq4$, with $n$ denoting the size of the approximation set. On the lower-bound side, exact value computation has an $\Omega(n\log n)$ lower bound in the algebraic decision-tree model already in two objectives, this bound lifts to every fixed $N\geq2$, and exact computation is $\#P$-hard when $N$ is part of the input. Together, the proposed perspective mapping provides a powerful tool for transferring algorithmic and structural results between anchored-box union and hypervolume theory and integral R2 computation.
arXiv:2607.01646v2 Announce Type: replace Abstract: State-of-the-art large language model (LLM) training takes tens of thousands of graphics processing units (GPUs) for months and encounters failures across the software and hardware stack. Existing fault-tolerance mechanisms either impose non-trivial overhead during failure-free execution or suffer from prolonged recovery latency, particularly under scenarios where a small subset of compute nodes experience permanent failures. %The tradeoff between failure-free overhead and recovery latency forms a space forms a Pareto frontier We present PHOENIX to simultaneously address both optimization objectives. PHOENIX incorporates a fault-tolerance mechanism that restores LLM training via hot-swapping, namely by replacing failed nodes with spare nodes without terminating the complete job. The hot-swapping of PHOENIX is enabled by two ideas: First, it exploits an off-critical-path in-memory checkpointing mechanism for spatial redundancy. Second, it introduces a communicator reconstruction protocol that replaces failed nodes with spare nodes at runtime. PHOENIX efficiently overlaps the in-memory checkpointing with computation, thus introducing zero overhead during error-free execution. Upon permanent node failures, PHOENIX can rebuild memory states with minimal recomputation by leveraging in-memory checkpoints. We evaluate PHOENIX across scales (up to 512 NVIDIA A100 GPUs) and LLMs (up to 65B parameters), and observe zero checkpoint overhead with hot-swapping recovery completing in under 40 seconds. These results show that PHOENIX simultaneously achieves both zero-overhead error-free execution and extremely low recovery cost.
arXiv:2607.01880v2 Announce Type: replace Abstract: Value functions are an essential component in actor-critic based deep reinforcement learning (RL). Conventionally, these functions are trained as a regression task by minimising the mean squared error (MSE) relative to bootstrapped target values. Meanwhile, in distributional RL, a distribution of returns is modelled based on the distributional Bellman operator. This work investigates the Gaussian Histogram Loss (HL-Gauss), a recent approach that reframes value estimation as classification by encoding each scalar Bellman target as a Gaussian-smoothed categorical target. Despite its potential, applying histogram-based losses to RL presents inherent challenges, most notably the requirement to pre-define a fixed support interval, which is often complicated by the non-stationary and stochastic nature of target values typically found in RL tasks. In this work, we propose an approach that dynamically learns the lower and upper bounds of the support instead of assigning them beforehand. We derive an objective that jointly learns these bounds whilst learning the categorical representation of the scalar values, and we show that this objective forms an upper bound on the mean-squared Bellman error. Our theoretical analysis further shows that this bound is tighter than that of non-learned supports of HL-Gauss. Empirically, the proposed objective enables stable adaptation of the support interval and matches HL-Gauss-based actor-critic algorithms on most continuous-control tasks whilst improving on a subset, without requiring a pre-specified support interval.
arXiv:2607.02205v2 Announce Type: replace Abstract: Sim-to-real transfer in robot learning is often limited by discrepancies between the ideal actuator dynamics assumed during policy training and the nonlinear, hardware-dependent behavior of physical motors. While conventional approaches attempt to bridge this gap by increasing simulator fidelity through system identification, domain randomization, or learned actuator models, we introduce an alternative paradigm: actuator reality shaping. Instead of modifying the simulator to match the real world, our method shapes the closed-loop behavior of physical actuators to match the idealized second-order reference dynamics used in simulation. By equipping each joint with a two-degree-of-freedom feedforward--feedback controller, we decouple reference-response shaping from robust stabilization, thereby providing a standardized actuator interface for reinforcement learning policies. As a result, policies trained only with the prescribed reference model can be deployed zero-shot on real hardware without task-level fine-tuning or learned actuator models. We validate the approach on a single-joint high-gear-ratio servo under external loads and a 7-DOF robotic arm reaching task, where actuator reality shaping substantially reduces sim-to-real tracking error and improves zero-shot task performance compared with standard servo-control and representative real-to-sim-to-real baselines. We further demonstrate zero-shot transfer on a wheeled-legged robot driving over a slope and a humanoid robot walking, suggesting that actuator reality shaping can serve as a reusable interface for robot learning across diverse hardware platforms. Project page: https://syamamori.github.io/ActuatorRealityShaping.github.io/
arXiv:2607.02275v2 Announce Type: replace Abstract: We present Cadence, a Byzantine fault-tolerant multi-proposer consensus protocol with arbitrarily low block intervals, optimal resilience, and optimal fast-path latency. Cadence divides time into equally spaced slots, one block per slot, each finalized in its own consensus instance. Blocks do not build directly on their predecessor, so instances run independently and none waits for an earlier block to finish or propagate; we call this extreme pipelining, decoupling the block interval from network latency. Cadence also removes the single-leader monopoly over transaction inclusion and ordering: under multiple concurrent proposers (MCP), several validators propose for each block, and it guarantees that, under synchrony, a transaction a correct proposer includes cannot be censored or deferred (short-term censorship resistance), and that no proposer can craft its proposal in reaction to the others' (hiding). To realize extreme pipelining, we introduce a general framework that turns any one-shot consensus meeting our slot-consensus specification into a multi-shot protocol. We instantiate it for MCP with two protocols of our own: Chorus, a slot consensus whose fast path finalizes a block in an optimal three rounds, with speculative finality one round earlier, and Conductor, an orchestrator that opens slots at an even cadence, more slowly under asynchrony to keep open slots bounded. To our knowledge, Cadence is the first MCP protocol to provide short-term censorship resistance and hiding at the fast-path latency of single-leader consensus. We prove safety, liveness, censorship resistance, and hiding under partial synchrony with optimal resilience (n = 3f+1). In simulation over Monad's 200 validators with five proposers per slot, finalization averages 219 ms (167 ms to speculative finality); at a 100 ms block interval a transaction waits on average 50 ms to enter a proposal.
arXiv:2607.04837v2 Announce Type: replace Abstract: Large-scale humanoid motion-tracking controllers are commonly improved by reallocating training effort: difficult motions are sampled more often, isolated into smaller subsets, or assigned to specialized experts. We show that this view is incomplete. In strong whole-body-control baselines, a residual set of feasible training clips remains unsolved even under targeted training, especially for high-dynamic transitions and balance-critical motions. These failures arise not only from insufficient exposure, but from a mismatch between the motion demands and the effective capability induced by the default training recipe. We propose Athena-WBC, a compact teacher-student pipeline with capability-aligned policy experts for long-tail humanoid whole-body control. Dynamic experts use a tracking-focused, constraint-aware objective that removes conservative effort and temporal-control penalties while preserving physical feasibility constraints; balance experts use a gravity curriculum to improve early-training survivability. The resulting privileged teachers are motion-routed for DAgger distillation and then compressed into a single controller with deployable observations followed by RL fine-tuning. Experiments on a full-size humanoid show improved recovery of training-set long-tail motions and better held-out tracking than a strong SONIC-recipe baseline, using only a small number of experts.
arXiv:2607.05001v2 Announce Type: replace Abstract: Cyber Threat Intelligence (CTI) reports are predominantly unstructured, heterogeneous, and noisy, which limits their direct usability for automated analysis and reasoning. Cybersecurity Knowledge Graphs (CSKGs) provide a structured representation of adversarial entities, actions, and relations, but constructing such graphs from free-text CTI remains a challenge. Recent approaches rely on monolithic Large Language Models (LLMs) to perform end-to-end extraction and completion, leading to high cost, limited controllability, and unstable performance. This paper introduces TACTIC-KG, an agentic framework for CSKG construction that decomposes the task into modular, specialized LLM agents responsible for extraction, typing, verification, and curation. Using lightweight models (3B--8B), TACTIC-KG improves stability, recall, and graph consistency while reducing deployment cost. We implement and evaluate TACTIC-KG against recent state-of-the-art systems. Experiments on human-annotated CTI reports show that agent specialization consistently outperforms larger monolithic in-context-learning (ICL) baselines in extraction F1-score, typing accuracy, and structural graph similarity.