Searcharxiv⌕ Search

arXiv subjects

Jun Wang

Publications and source records attributed to Jun Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

MILO: Efficient Many-shot In-Context Learning with Block-wise Low-rank Compression

Many-shot in-context learning (ICL) enables large language models (LLMs) to adapt to complex tasks by conditioning on thousands of demonstration examples, but this paradigm shifts the inference efficiency bottleneck to the key-value (KV) cache memory. Due to the linear scaling behavior of the KV cache, storing these intermediate tensors has become a paramount challenge for both online serving and on-device deployment. To address this issue, we propose a novel compression framework, termed MILO, that exploits the low-rank redundancy inherent in many-shot contexts. Specifically, MILO features a block-wise low-rank compression strategy that compresses the KV cache at the block granularity, where each block contains multiple many-shot examples. Furthermore, to handle the heterogeneous context density across different blocks, MILO dynamically allocates rank budgets based on the information entropy, preserving the fidelity of critical blocks while aggressively compressing redundant ones. Experimental results on Qwen2.5 models demonstrate that our method achieves up to 50% reduction in KV cache memory and 1.8x throughput improvement, with negligible performance degradation on classification and reasoning benchmarks, significantly outperforming prior baselines.

cs.CL↗
arXiv2026

What Makes an Efficient VLA? Navigating Action-Head Design, Scaling, and Latency

Vision-Language-Action (VLA) models combine a pretrained vision encoder, a language backbone, and an action head, but their relative contribution has not been established under controlled, latency-paired conditions. We fix the backbone families (SigLIP2 and Qwen2.5) and the training pipeline, sweep action-head design and module scale, and pair each configuration with measured on-device latency. The study yields three findings. First, action-head performance is governed primarily by initialization rather than decoder architecture, loss, or inference budget: copying the last transformer layers of the language backbone into the head is the single largest lever, at no latency cost, and the only axis that helps at every module scale. Alignment also explains the other axes: flow matching and a heavier decoder pay off only while the head is misaligned and reverse once it is aligned, and extra inference passes give no measurable benefit; expressiveness appears to substitute for missing alignment. We read this as representation transfer: the aligned head keeps attending to the instruction's object nouns and stays close to the backbone in weight space rather than relearning to act from scratch. Because we reach alignment only through initialization, we offer this as the account that best organizes the measurements, not a demonstrated cause, and name the control that would settle it. Second, capacity pays only after alignment: the aligned action head is the highest-return module to scale. Third, those returns diminish sharply near the size today's $π$-series VLAs already use, so further growth buys little in-domain accuracy for its latency. These specify EffVLA, a compact model matching the strongest open-source VLAs on standard LIBERO, leading on most LIBERO-Plus perturbation axes at lower latency, and transferring to a real SO-ARM101 arm with the recipe unchanged.

cs.RO↗
arXiv2026

Thermodynamics of arbitrary spherical surfaces in FLRW spacetimes

In this work, we investigate the thermodynamics of arbitrary spherically symmetric surfaces in non-flat FLRW spacetimes within general relativity. Adopting the gravity-to-thermodynamics perspective, the Clausius relation and the unified first law follow directly from the gravitational field equations and matter dynamics, with a central role played by the matter-flux measure derived from energy-momentum conservation. The cosmological apparent-horizon and event-horizon thermodynamics are both recovered as special cases. In particular, the event-horizon temperatures derived from the gravitational dynamics satisfy both thermodynamic relations without introducing phenomenological modifications, and the corresponding temperature reduces to the Gibbons-Hawking temperature in the de Sitter limit. We further show that previous inconsistencies originate from incompatible temperature prescriptions rather than from any intrinsic deficiency in the thermodynamics of the cosmological event horizon. These results demonstrate that the gravity-thermodynamics correspondence is a generic feature of gravitational dynamics rather than a property unique to causal horizons.

gr-qc↗
arXiv2026

Information-Time Proximal Policy Optimization

RLVR has substantially improved the reasoning capabilities of LLMs. However, existing methods typically parameterize temporal progression in the Markov Decision Process by token-by-token generation, despite the highly non-uniform information flow along autoregressive trajectories. In this paper, we propose InfoPPO, which reparameterizes temporal progression using information density rather than raw token count. This reparameterization induces a common state-dependent structure for both temporal credit propagation and policy updates. InfoPPO restores the effectiveness of non-trivial discounting in long-horizon reasoning, retaining effective-horizon contraction while avoiding excessive attenuation of terminal supervision over long token sequences. Moreover, the information-time policy-improvement analysis naturally leads to a state-dependent update constraint, which we implement through adaptive clipping. By adapting the clipping threshold at each token position to the information density of its corresponding state, this mechanism enables more targeted policy updates while preserving proximal control. Theoretically, we extend performance-difference and policy-improvement analyses to the information-time MDP, deriving a policy-improvement lower bound when policy changes are regulated by information density. We further connect the general information-time analysis to practical LLM policy optimization by relating state-wise information density to local policy movement, while also providing theoretical grounding for the adaptive update mechanism. Experiments on Qwen3 models demonstrate consistent gains over competitive baselines across five challenging competition-style mathematical reasoning benchmarks. InfoPPO also maintains stable accuracy and response length across non-trivial discount settings under which token-time PPO deteriorates.

cs.LG↗
arXiv2026

R3D: Revisiting 3D Policy Learning

3D policy learning promises superior generalization and cross-embodiment transfer, but progress has been hindered by training instabilities and severe overfitting, precluding the adoption of powerful 3D perception models. In this work, we systematically diagnose these failures, identifying the omission of 3D data augmentation and the adverse effects of Batch Normalization as primary causes. We propose a new architecture coupling a scalable transformer-based 3D encoder with a diffusion decoder, engineered specifically for stability at scale and designed to leverage large-scale pre-training. Our approach significantly outperforms state-of-the-art 3D baselines on challenging manipulation benchmarks, establishing a new and robust foundation for scalable 3D imitation learning. Project Page: https://r3d-policy.github.io/

cs.CV↗
arXiv2026

Shake to Learn: Dynamic Interrogation of Hidden Object Physics for Robotic Manipulation with Physical Reservoir Computing

Many physical properties relevant to robotic manipulation are hidden from vision. A sealed object, for example, may reveal little about its center of mass (COM) or internal contents until it is lifted, shaken, or otherwise dynamically perturbed. This study shows that such interactions can enable a new modality of robotic perception and learning, in which interaction-induced dynamic responses are used to infer object physics that is inaccessible to conventional sensing. We implement this idea using an origami-inspired soft robotic arm that functions as a physical reservoir computer. After grasping an object, the arm is excited by a fixed shaking input at its base, and the resulting ringdown response is recorded through either camera tracking or embedded sensors. Because the input is held constant across trials, hidden object properties, such as the COM position, are encoded through their effect on the dynamics of the coupled robot-object system. A lightweight linear readout can then decode these dynamics to recover interpretable information about the hidden object physics. Using this framework, the soft robotic arm reservoir completed three tasks of increasing difficulty: inferring the orientation of the object's hidden COM, inferring the COM distance from the grasp point, and using the inferred COM information to guide a subsequent regrasp. We further develop a dynamic summary representation of the ringdown response that improves prediction accuracy. Together, these results establish shake-to-learn mechanical interrogation as a promising strategy for robotic systems to convert brief physical interactions into actionable cues about hidden object properties for downstream manipulation.

cs.RO↗
arXiv2026

JewelTry: Mask-Free Scale Aware Jewelry Virtual Try-On

Virtual try-on (VTON) enables customers to visualize how fashion products appear when worn and has become an important technology for online shopping. While recent advances have substantially improved garment VTON, jewelry remains a challenging and underexplored category due to its small size, rigid structure, and sensitivity to fine-grained visual details. Realistic jewelry VTON requires not only faithful appearance transfer but also accurate scale and placement relative to the wearer. Existing jewelry VTON methods typically rely on mask guidance, whereas mask-free approaches lack explicit guidance for modeling the product scale. To bridge this gap, we introduce JVTO-Bench, a benchmark dataset for scale-faithful jewelry VTON, providing reference source target triplets with real-world product-scale annotations across four major jewelry categories. Building upon this benchmark, we propose JewelTry, a mask-free diffusion framework for scale-aware jewelry VTON. JewelTry incorporates a scale adapter that encodes product dimensions into a scale token, enabling the model to learn scale relationships between jewelry items and surrounding human anatomy in-context. To further improve jewelry consistency, we introduce a single-directional condition attention mechanism and an attention refinement loss that preserve both coarse geometry and fine-grained structural details of the reference jewelry. Extensive experiments show that JewelTry achieves a balance among visual fidelity, background preservation, object consistency and scale accuracy, establishing a strong baseline for mask-free, scale-aware jewelry virtual try-on.

cs.CV↗
arXiv2026

Embracing Flow Unsteadiness: A High-Throughput Learning Platform Enables Vortex-Exploiting Bioinspired Propulsion

Biological swimmers and flyers exploit unsteady vortices for propulsion, whereas engineered vehicles usually suppress them as disturbances. Learning such flow exploitation in machines is difficult because real-fluid interaction data are scarce and unstructured exploration is unstable in high-dimensional, history-dependent flows. Here we present REEF, a co-designed physical-learning framework that integrates SHOAL, an eight-channel high-throughput array for real fluid--structure interaction, with V-STAR, a staged algorithm that converts these interactions into policies through imitation, offline internalization, and online adaptation. Across lift-based, drag-based, and momentum-jet propulsors, REEF expands the attainable force envelope to more than twice that of parameterized search. Particle image velocimetry shows that these gains arise from coordinated vortex formation, growth, and force projection, rather than refinement of a fixed motion-to-force mapping. Force-trained policies transfer zero-shot to free-moving robots whose body motion changes the surrounding flow, suggesting that REEF learns transferable wake-coupling principles for embodied propulsion in unsteady fluids.

physics.flu-dyn↗
arXiv2026

A property-registry contract for retrieve-or-refuse thermal-mechanical lattice search

Early thermal-mechanical lattice requirements are knowledge-intensive and often jointly unsatisfiable: an engineer asks for a cell that is light, stiff, laterally conducting and cheap, and no cell in the library satisfies it. A design system should say so, and say which requirement to loosen and by how much, rather than return the nearest row. A generative model can return a candidate even when the library holds none. This work applies established conflict diagnosis to a catalogue of homogenised properties. Search over 1,397 homogenised unit cells and 19 base materials (26,543 combinations) returns a catalogue row that a second solver can rebuild, or, for an empty feasible set of at most eight constraints, inclusion-minimal unsatisfiable subsets (MUS) and the slack of a repair, not a neighbour. One registry declaration generates the prompt and the evaluator, so undeclared parsed keys cannot reach search. On a frozen suite of 64 typed queries, retrieve-or-refuse matches min-repair on every feasible query (48/48) and refuses every empty one (16/16) with MUS and slack. Constraint-ignoring nearest-neighbour and penalty search violate stated cost or density limits. On 216 empty queries, repairs printed at three significant figures and rounded outward stay feasible in every case (216/216). When density and cost must be kept, a list of minimum-cardinality repairs keeps them on 209 of 211 queries; a protection-first repair keeps them on all 211, the same as full diagnosis. A 308-request parse benchmark is a template-text check, not unconstrained engineer prose.

cs.CE↗
arXiv2026

Retrieval-Driven Memory Reconsolidation for Long-Term LLM Agents

Long-term memory is essential for LLM-based agents operating over extended interactions. Existing memory systems primarily update memory when new information arrives, treating retrieval as the endpoint of memory access rather than a driver of memory evolution. Consequently, retrieval feedback is rarely exploited to reorganize memory for future access continuously. Moreover, most existing approaches rely on predefined memory structures together with fixed retrieval pipelines, limiting the agent's ability to organize and evolve its own memory autonomously. Inspired by memory reconsolidation in cognitive neuroscience, we propose \textbf{REALM}, a \textbf{r}econsolidation-\textbf{e}volution \textbf{a}gentic \textbf{l}ong-term \textbf{m}emory framework. It models long-term memory as a continual lifecycle by autonomously organizing memories into a heterogeneous cognitive graph, retrieving evidence via adaptively composed graph-search atoms, and continually reconsolidating memories based on retrieval feedback. REALM achieves an average accuracy of 75.97\% on LoCoMo and 65.11\% on LongMemEval, outperforming the strongest baselines by 7.17 and 1.31 points respectively. Ablation studies confirm that memory reconsolidation consistently boosts performance, with further analyses revealing that it progressively reorganizes related memory units into more coherent local structures for collective evidence recall and utilization during reasoning. These results suggest that retrieval-driven memory reconsolidation provides an effective mechanism for continually evolving long-term memory in LLM agents.

cs.CL↗
arXiv2026

PRISM-Bench: An Audio-Centric Diagnostic Benchmark for Text-to-Audio-Video Generation

Text-to-audio-video (T2AV) generation has advanced rapidly, but its evaluation still underestimates the audio modality. Existing benchmarks either treat audio as an auxiliary component of video quality or assess it in isolation from audiovisual grounding, making it difficult to diagnose where current systems truly succeed or fail in audio generation. We present PRISM-Bench, the first audio-centric diagnostic benchmark for T2AV generation. Built from a rigorously curated dataset of 900 human-verified samples, PRISM-Bench factorizes audio evaluation along two orthogonal axes: audio type (Speech, Music, and Sound) and sound-source visibility (On-screen vs. Off-screen). It evaluates generated content across four perceptual dimensions (Audio-Visual Coherence, Audio Quality, Audio Expressiveness, and Prompt Following) with 35 fine-grained criteria. To ensure reliable assessment, we adopt an enhanced MLLM-as-a-Judge protocol based on blind, side-by-side comparison against ground-truth references, demonstrating strong alignment (over 70% mean agreement) with human raters. Our evaluation of recent T2AV systems highlights a significant performance gap between frontier and open-source models. Furthermore, we demonstrate that current generation paradigms overfit to perceptual fidelity while struggling with complex grounding and control tasks, particularly in generating music and synchronized On-screen audio.

cs.MM↗
arXiv2026

Safety in Self-Evolving Agents: A Survey

Large language models (LLMs) exhibit strong general capabilities, yet their parameters typically remain fixed after deployment, limiting learning from new interactions. In open-ended environments, this motivates self-evolving agents that continually update reusable state-including model parameters, memories, tool definitions, skills, and workflows-from data, feedback, and accumulated experience. This shift changes the safety problem: once experience becomes reusable state, past events become future causes, and information harmless in one context may later influence decisions with greater persistence, authority, or scope. Self-evolving agent safety therefore asks not only whether a response is aligned or an action authorized, but whether safety properties survive the accumulation, generalization, and cross-context reuse of locally useful experience. We introduce SAVER, a transition-centered framework in which Substrate locates reusable influence, Adaptation captures how it changes, Violation identifies compromised safety attributes, Exposure marks where failures become observable, and Response assesses containment, repair, or revocation. Our survey reveals that failures need not originate from harmful information: legitimate state can become unsafe when adaptation expands its persistence, authority, or scope beyond the conditions under which it was valid. Existing work provides comparatively strong evidence for admission, retrieval, activation, exposure, and local containment, but much less for descendant repair and evaluation after adaptation resumes. We therefore argue for longitudinal evaluation that traces unsafe influence to its originating transition, verifies repair across descendants, and tests whether it can re-emerge under continued evolution.

cs.CR↗
arXiv2026

Skeleton-Guided Sparse Anchors for Rotated Instance Segmentation in Cell Microscopy

One of the fundamental challenges in cell microscopy (MS) image analysis is instance segmentation (IS), particularly when segmenting cluster regions where multiple objects of varying sizes and shapes may be connected or even overlapped in arbitrary orientations. Existing IS methods usually fail in handling such scenarios, as they rely on coarse instance representations such as keypoints and horizontal bounding boxes (h-bboxes). In this paper, we propose a novel one-stage framework named A2B-IS to address this challenge and enhance the accuracy of IS in MS images. Our approach represents each instance with a pixel-level mask map and a rotated bounding box (r-bbox). Unlike two-stage methods that use box proposals for segmentations, our method decouples mask and box predictions, enabling simultaneous processing to streamline the model pipeline. Additionally, we introduce a Gaussian skeleton map to aid the IS task in two key ways: (1) It guides anchor placement, reducing computational costs while improving the model's capacity to learn RoI-aware features by filtering out noise from background regions. (2) It ensures accurate isolation of densely packed instances by rectifying erroneous box predictions near instance boundaries. To further enhance the performance, we integrate two modules into the framework: (1) An Atrous Attention Block (A2B) designed to extract high-resolution feature maps with fine-grained multiscale information, and (2) A Semi-Supervised Learning (SSL) strategy that leverages both labeled and unlabeled images for model training. Our method has been thoroughly validated on two large-scale MS datasets, demonstrating its superiority over most state-of-the-art approaches.

cs.CV↗
arXiv2026

Nonlinear Dissipation and Hopf Criticality in Driven Dissipative Collective Spins

Self-sustained oscillations combine finite-amplitude stabilization with a neutral phase degree of freedom. We develop this bifurcation-based framework for driven-dissipative collective spins and show that the microscopic structure of the U(1)-covariant dissipation selects the background attractor, while the explicit U(1)-breaking channel governs its local bifurcation response. In the thermodynamic-limit mean-field dynamics, a single linear U(1)-covariant jump produces only polar fixed-point backgrounds, whereas nonlinear covariant dissipation provides amplitude-dependent saturation and stabilizes a finite-latitude self-sustained-oscillator manifold through a supercritical Hopf bifurcation. Under coherent U(1) breaking, exact resonance leads to a reversible double-zero degeneracy with vanishing critical frequency rather than a standard Hopf onset. Finite detuning unfolds this singularity into a genuine finite-frequency Hopf boundary, which exists only on the self-sustained-oscillator side and can be either supercritical or subcritical. By contrast, a single linear dissipative U(1)-breaking jump cannot generate a standard Hopf instability: when its phase-pinning invariant vanishes the azimuthal direction remains neutral, whereas otherwise the phase-locked fixed points have a purely real Jacobian spectrum. These results establish a general design principle: nonlinear covariant dissipation selects the selfsustained background, while the structure of the symmetry-breaking channel determines whether the resulting local response is double-zero, genuinely Hopf, or non-Hopf.

quant-ph↗
arXiv2026

Mutually phase-stable tunable attosecond soft X-ray attosecond pulses from a free-electron laser

We demonstrate the production of mutually phase-stable attosecond X-ray pulse pairs with tunable relative time delays and phases in a cascaded X-ray free-electron laser. We showcase the method in an experiment at the LCLS-II, in which a shaped electron beam is used in a split undulator configuration to generate the two attosecond pulses. We achieve mutual phase stability by reusing microbunching generated in the first undulator in order to seed the FEL process in the second at a detuned frequency. We measure controllable temporal delays between the two pulses directly in the time domain using angular streaking of photoelectrons, with a step size of 250 attoseconds. We then show that the behavior of the X-ray spectrum is consistent with phase stability between the two pulses, with a relative phase that can be easily tuned using inter-undulator phase shifters. This method is particularly well-suited to few to ten eV energy separations and sub to few femtosecond time delays, which are ideal for experiments in the soft X-ray regime for pushing the limits of our models for molecular dynamics and exerting direct coherent control over quantum systems.

physics.acc-ph↗
arXiv2026

SafeRestore: Detector-Relative Risk Certificates for Selective Industrial Image Restoration

Industrial inspection pipelines often restore a measured image before a detector acts on it, yet restoration can suppress detector-supported defect structure or create clean-region activations. We formulate restoration as a selective action problem over the measured display, five restored candidates, and review. SafeRestore ranks candidates with action-specific fitted scores, chooses a gate on threshold-tuning data, and evaluates the fixed gate on a disjoint certification sample with two one-sided exact binomial bounds: one for the positive-conditional evidence-loss incident rate and one for the all-accepted excess-activation incident rate. The guarantee is marginal for one policy fixed before its certification outcomes are observed, under an image-level i.i.d. working model. In a retrospective split-sample study of 4,591 public Carinthia-S images, the protocol yields auditable risk-coverage behavior. The primary all-action policy passes in one of five training repetitions (12.0% +/- 26.9% pass-gated test coverage when failures count as zero), whereas fixed bicubic and reduced-complexity variants pass more often. On reserved morphologies, evidence-loss incidence rises to 81.1-90.3%, and KolektorSDD lacks both detector competence and enough positive certification images for the stated target. The contribution is therefore an auditable, detector-relative framework for deciding when a transformed image may be returned automatically and when review remains necessary -- not a claim that adaptive routing outperforms simpler policies on the present evidence.

cs.CV↗
arXiv2026

Dynamical (non-)recurrence of escaping exponential maps

We show that exponential maps could have complicated measurable behaviours by giving two examples in the exponential family both with singular value escaping to infinity ``slowly", one of which is recurrent (in the sense that every positive measure set will return to itself infinitely many times), while the other is non-recurrent. This also complements earlier results of Lyubich, Rees, Urbański-Zdunik, and Hemke. We also give an example of exponential maps with singular value escaping to infinity arbitrarily slow, which is a parametric analog of a result of Rempe.

math.DS↗
arXiv2026

Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation

Multi-agent debate (MAD) improves the reasoning capabilities of large language models by having multiple agents iteratively refine their responses through discussion. However, MAD suffers from a critical vulnerability known as shared misconception: when a majority of agents initially converge on an incorrect answer, the debate process tends to amplify rather than correct the error. Existing methods primarily address peer skew but leave the agents' inherently biased concept priors unaddressed. To mitigate this systematic weakness, we propose R$^2$-MAD (Remember and Reweight for Multi-Agent Debate), a framework that equips agents with an experience memory accumulated from past debates. R$^2$-MAD intervenes on both failure modes through two complementary mechanisms: A debate-state-aware retrieval policy dynamically calibrates the concept prior by retrieving relevant historical evidence based on the current consensus level. Then these retrieved experiences provide a basis for estimating per-agent reliability, yielding confidence weights to modulate peer influence. Experiments on various benchmarks show that R$^2$-MAD achieves consistent improvements over existing single-agent and MAD baselines.

cs.CL↗