Computer Vision
DPAMixerSR: An Efficient Degradation-Pattern-Aware Model for Image Super-Resolution
Song-Li Wu, Haonan Jiang, Jixuan Fan et al. (+3 more)
While content-adaptive schemes have delivered notable advances in image super-resolution (SR), existing approaches typically focus on texture complexity and ignore intrinsic degradation factors (e.g., blur kernels or noise patterns), leading to suboptimal computation allocation and reconstruction performance. To remedy this, we propose DPAMixerSR, a degradation-pattern-aware framework that enables efficient SR through adaptive sparse computation. We design a lightweight Perceptual Degradation Ranking (PDR) module partitions the image into severely and mildly degraded patches, which are routed to the Adaptive Sparse Processing (ASP) and a lightweight convolutional branch, respectively. ASP performs structure-aligned, multi-scale sparse propagation and bidirectional refinement, while the convolutional branch enhances efficiency in mildly degraded regions. By coupling degradation-driven routing with structure-aligned sparse processing, DPAMixerSR establishes a self-regulating framework that dynamically balances computational efficiency and reconstruction fidelity. Extensive experiments on various SR tasks demonstrate that our DPAMixerSR achieves superior structural restoration and perceptual fidelity with markedly reduced computational overhead, providing a novel and scalable framework for degradation-aware, resource-efficient SR.
Read the full abstract ▾
All authors Song-Li Wu, Haonan Jiang, Jixuan Fan, Yufei Huo, Chubin Zhang, Yansong Tang
Artificial Intelligence
CoWindow Attention: Full Causal Coverage Is a Collective Property
Jingze Shi, Zhangyang Peng, Xianduo Li et al. (+6 more)
FullAttn repeatedly exposes the complete causal history to every
attention head, creating substantial redundant computation and
memory traffic even with IO-efficient dense kernels. We introduce CoWA, a structured
attention architecture that distributes access to the causal history across KV heads. All heads share near-diagonal and prefix-sink windows, while complementary long-range windows partition the remaining history. Their union provides full causal coverage although each head attends sparsely to distant
tokens. This position-defined
attention pattern requires no learned router or indexer, is used consistently during training and inference, and aligns with KV-head tensor parallelism. A window-matched ablation at 8K isolates the effect of complementary long-range allocation: CoWA with 100% collective coverage reaches 89.73% accuracy, compared with 89.97% for FullAttn, while duplicated long-range windows perform substantially worse. Across a broader controlled associative-recall comparison with matched
token budgets, CoWA closely tracks FullAttn as the context grows, whereas other sparse patterns lose a substantial fraction of the associations. In an
attention-operator
benchmark at 128K
tokens with tensor parallelism, CoWA reduces forward and backward latency during training by 7.4x and 8.6x and decoding latency during inference by 3.0x over FullAttn. Its per-rank peak operator
memory matches FullAttn during training and is 7.6x lower during decoding. Across scaling-law training from 0.6B to 14B parameters, CoWA closely tracks FullAttn in perplexity while reducing total training FLOPs. The resulting 14B models and 32B models from separate continued training achieve comparable knowledge,
reasoning, and long-context
retrieval scores to FullAttn. These results show that full causal coverage can be a collective property of the head ensemble rather than a duplicated property of every head.
Read the full abstract ▾
All authors Jingze Shi, Zhangyang Peng, Xianduo Li, Yanlin Qi, Xiaotian Lin, Haoxian Chen, Liangdong Wang, Guang Liu, Yuyu Luo
Dynamical Systems
Non-analyticity of hairs of exponential maps
Weiwei Cui, Jiaxing Huang and Lingrui Wang
Let $f_{\lambda}(z)=\lambda e^z$, where $0<\lambda<1/e$. In 1984, Devaney and Krych showed that the Julia set consists of pairwise disjoint hairs. Viana further in 1988 proved that these hairs are $C^{\infty}$. In this paper we show that hairs are not analytic except for trivial ones. This solves a long-standing open question.
Read the full abstract ▾
High Energy Physics (Theory)
Greybody Factors and Hawking Spectra of Quantum-Corrected Black Holes
Moslem Shafiee and Ahmad Sheykhi
We investigate Hawking radiation of a
quantum-corrected Schwarzschild black hole with
vacuum polarization and conformal anomaly effects. The corrected geometry has two horizons, a non-monotonic Hawking temperature that vanishes at a critical mass, and a logarithmic
entropy correction. We compute greybody factors, spectra, and emission rates for massless bosonic fields of various
spins and angular momenta.
Quantum corrections suppress low-energy greybody factors and enhance high-energy ones. During late evaporation, where
quantum influences become significant, the decreasing temperature strongly suppresses emission. In this case, the higher-
spin modes become negligible, and the scalar $s=l=0$ mode dominates. The black hole lifetime is greatly extended as it asymptotically approaches a stable extremal remnant.
Read the full abstract ▾
Artificial Intelligence
Despite Instructions: Frontier Agents Improvise Covert Channels at Test Time
Jacob Dineen, Silei Ren, Muhao Chen et al. (+2 more)
In security-sensitive applications,
language-model
agents are often required to coordinate without disclosing confidential information. Yet repeated interactions may also let ordinary messages acquire shared private meaning. We study a repeated game with pairs of models in which the sender model observes one of four secret states and selects one of four summaries of the same public report, while the receiver model tries to infer the secret state. We find that model pairs can learn to communicate the secret using only one bit of feedback indicating whether the receiver inferred it correctly. This learning occurs during inference with fixed parameters and no supplied codebook or encoding examples. The effect also persists when
agents generate their own free-form updates in a simulated incident-response task. Across ten independent games, pairs of GPT-5.6 Sol
agents reach 98.8% final accuracy, compared with 25% chance, despite explicit instructions prohibiting disclosure and a monitor that screens each message without access to the
agents' interaction histories. The same interactions that help
agents cooperate can therefore allow confidential information to pass through messages intended for legitimate coordination.
Read the full abstract ▾
All authors Jacob Dineen, Silei Ren, Muhao Chen, Dan Roth, Ben Zhou
Artificial Intelligence
CAIRN: Dynamic Fact-Intent DAGs for Multi-Agent Exploration
Zuyao Xu, Yuyang Jia, Junwei Guan et al. (+3 more)
LLM-powered autonomous systems have demonstrated promising capabilities in mathematical
reasoning, engineering, and cybersecurity. Yet how to organize these systems for effective, reliable, and sustained performance remains an open question. In this paper, we present CAIRN, a fact-intent-driven multi-
agent paradigm for goal-directed exploration. CAIRN represents observations and planned investigations as a dynamic directed acyclic graph (DAG). A reasoner interprets facts to propose intents, which workers execute to produce new facts. Each intent references its supporting facts and defines a potential exploration branch. The persistent graph preserves goals, dependencies and findings across workers, supporting knowledge reuse and parallel exploration. The graph also makes execution trajectories traceable and auditable, providing a basis for human verification and intervention. We evaluate CAIRN across cybersecurity and mathematical
reasoning tasks, examining task success, time to solution, and
token consumption. DAG-based coordination can incur higher
token costs with no observable performance gains on tasks that require little effort. However, on high-effort tasks (at least 1M
tokens), we observe faster solutions in 76.5% of cases, with speedups of up to 3.08x. Moreover, as task effort increases, these time gains become more pronounced while relative
token overhead declines, highlighting the potential of DAG-guided parallel exploration.
Read the full abstract ▾
All authors Zuyao Xu, Yuyang Jia, Junwei Guan, Xiang Li, Kaiwen Shen, Zhiqiang Dong
Information Retrieval
RandSlot: Learning Compact Visual Document Representations with Random Soft Tokens
Dewen Guo, Shi Yu, Lingxiao Zhang et al. (+3 more)
Visual document
retrieval requires expressive representations to match queries with evidence distributed across text, tables, and page layouts. Multi-vector representations capture fine-grained information, but storing and comparing many vectors introduces substantial
retrieval costs. In this paper, we introduce RandSlot, a simple approach to learn compact visual document representations with random soft
tokens. During training, we append independently-sampled random unit vectors to query and document input sequences and resample them at every use, without introducing learnable soft-
token parameters. The encoder contextualizes these auxiliary inputs with the original content to produce a small set of
retrieval vectors. A standard late-interaction objective trains the encoder to extract relevant information under varying input conditions. Experiments with different backbone models show that RandSlot improves
retrieval quality over alternative readout strategies under the same vector budget. Further analysis shows that these gains can persist when random soft
tokens are replaced with zeros at inference, demonstrating that random inputs during training can improve compact
retrieval representations even when inference no longer requires sampling.
Read the full abstract ▾
All authors Dewen Guo, Shi Yu, Lingxiao Zhang, Yang Zhang, Tao XU, Dan Wang
Robotics
SEES: A Self-Evolving Embodied System via Failure-Guided VLA Policy Adaptation
Ziwen Li, Hanlue Zhang, Zhenyang Ren et al. (+9 more)
Recent
vision-
language-action (VLA) policies demonstrate promising generalization across diverse short-horizon tasks. However, they remain unreliable on long-horizon tasks, partly because the large-scale training data is biased toward single-stage manipulation tasks that are cheaper to demonstrate. A single weak atomic skill can cause failures across multiple multi-stage tasks. To address such failures, existing methods often require experts to identify the bottleneck and provide additional demonstrations, making the improvement costly and potentially impractical after deployment. To this end, we present a Self-Evolving
Embodied System (SEES) that learns from failures and improves the VLA policy without additional expert demonstrations. SEES decomposes long-horizon tasks into atomic tasks and routes them to corresponding family policies. Each family consists of related atomic skills that share one VLA adapter. During execution, the system automatically monitors atomic-task outcomes to identify the most frequently failing atomic skills as the current bottlenecks. To overcome these bottlenecks, SEES constructs tailored RL tasks in simulation by restoring previously encountered states and generating task-specific success criteria with an LLM. Online RL updates the shared family adapters to promote positive transfer among related atomic skills and cumulative improvement across
evolution rounds. Extensive experiments show that SEES can be integrated with different VLA backbones to progressively improve their long-horizon performance. We also observe continued improvement on unseen tasks, providing evidence of transfer beyond the
evolution settings.
Read the full abstract ▾
All authors Ziwen Li, Hanlue Zhang, Zhenyang Ren, Tianyu Huang, Runqi Lin, Haoyu Wang, Zhengqing Gao, Yandong Guo, Fakhri Karray, Tongliang Liu, Chris Russell, Mingming Gong
Classical Analysis and ODEs
Level set estimates for strictly convex and hyperbolic functions
Xianghong Chen and Tongou Yang
In this note, we prove uniform upper bounds for the volume of the level set $$\{x\in\Omega: c\le f(x)<c+\delta\},\quad c\in\mathbb R,\;\delta>0,$$ for strictly convex and
hyperbolic functions $f$ defined on a convex domain $\Omega\subset\mathbb{R}^n$ ($n\ge 2$). In particular, under a Hessian lower bound $D^2f\ge I_n$, we obtain the sharp volume bound $$|S^{n-1}|\, r(\Omega)^{n-2}\,\delta,$$ where $r(\Omega)=\frac12\,{\mathrm{diam}\Omega}$. As an application, we derive $L^2$ estimates for oscillatory integrals with convex phases.
Read the full abstract ▾
Artificial Intelligence
Flat-Consensus Diffusion for Robust Data Reshaping under Noisy Evaluator
Hongyu Cao, Kunpeng Liu, Fei Xie et al. (+1 more)
Data shape determines how features are structured, how patterns are separated, and how distributions cover the underlying domain. Poor data shape can make models learn noise rather than generalizable structure. This paper studies robust feature-centric data reshaping: generating feature transformations that remain useful, stable, and reproducible under noisy evaluation and imperfect data conditions. We view reshaping operation sequence search as reward-guided
diffusion generation, and robust reshaping as searching for regions in the latent reward landscape rather than isolated high-reward transformations. The key challenge is dual instability: noisy evaluators distort local reward guidance, while
stochastic generative trajectories can converge to inconsistent solutions. We propose FCDiff, a flat-consensus
diffusion framework that addresses both failures through a micro-macro decomposition. The micro layer replaces point-estimate reward guidance with Gaussian-smoothed, Monte Carlo averaged gradients, steering generation toward locally flat reward regions. The macro layer aggregates independently guided trajectories with a weighted Frechet-mean barycenter, selecting consensus-supported basins and filtering
stochastic outliers. Across an 8-
dataset headline cohort under heavy-tailed evaluator noise, FCDiff attains the best aggregate rank on lower-tail reliability and robustness against both search-based AutoFE and robustness-oriented generative baselines, with statistically significant accuracy gains over every generative baseline. Our results show that robust data reshaping requires searching for flat, consensus-supported regions rather than sharp single-trajectory optima.
Read the full abstract ▾
All authors Hongyu Cao, Kunpeng Liu, Fei Xie, Sandip Ray
Machine Learning (Statistics)
Affine Geometry of Gaussian ReLU Networks via Conditional Kac-Rice Formulas
Recep Özkan and Christian Hirsch
We study how the affine geometry of finite ReLU networks is created at random initialization and reorganized by supervised training. We call a sign-changing zero of a hidden preactivation an activation switch and a point where the scalar network output is nondifferentiable a scalar kink. For one-dimensional input, conditioning on the preceding layers makes each preactivation Gaussian and affine on the
cells of a random finite partition. This yields an exact finite-width conditional Kac-Rice formula for the expected number of activation switches along an input interval. For fixed depth and proportionally growing widths, the resulting switch intensities converge to explicit deterministic limits. A visibility estimate shows that the expected number of switches that do not produce scalar kinks is negligible, yielding an explicit leading formula for the expected number of scalar kinks and hence affine regions. In higher input dimensions d >= 2, the analogous conditional surface formula yields the leading expected (d-1)-dimensional Hausdorff measure of the scalar kink set. On the Breast Cancer Wisconsin data, the initialization formula accurately predicts switch counts along held-out segments. After training, switch counts decrease along within-class segments and increase along between-class segments. Thus, training redistributes rather than merely contracts affine complexity.
Read the full abstract ▾
Artificial Intelligence
IGSD: Environment-Verified Hindsight Self-Distillation for Search Agents
Angqing Jiang, Gaoming Zhang, Chaoqun Zhang et al. (+5 more)
On-policy self-
distillation densifies
agent training without external teachers: a policy conditioned on privileged hindsight provides step-level guidance for its own unprivileged rollouts. For search
agents, however, hindsight can make the teacher prefer a query that does not improve
retrieval from the student's state. Existing methods either distill this preference directly or filter it with model-internal scores, but neither strategy verifies the query's executed
retrieval consequence. We propose Information-Gain-Gated Self-
Distillation (IGSD), which verifies on-policy
token proposals with environment feedback before distilling them. Treating each query
token as a micro-action, IGSD completes the teacher's
token proposal and the student's sampled
token into matched queries and executes both from the same failed state with the same retriever. Shared counterfactual controls account for query-conditioned shifts in answer likelihood, so their difference, the executed paired information gain, provides a relative utility contrast for the retrieved documents. IGSD uses this contrast as a positive-only soft weight for candidate-pair
distillation, while leaving the GRPO objective unchanged and confining verification to training. Across seven single-hop and multi-hop QA
benchmarks, IGSD reaches macro-average exact-match accuracies of 42.8% and 47.0% with 3B and 7B policies, respectively, without inference-time verification. These results support environment-verified hindsight as an effective approach to reliable action-level supervision for search
agents.
Read the full abstract ▾
All authors Angqing Jiang, Gaoming Zhang, Chaoqun Zhang, Jianchun Song, Liyuan Kong, Kena Qi, Wei Lin, Defu Lian
Analysis of PDEs
Singular analysis of the principal eigenvalue for cooperative elliptic systems with drifts
Shuang Liu
We determine the small-
diffusion limit of the principal
eigenvalue for a one-dimensional cooperative
elliptic system with component-dependent drifts and Neumann boundary conditions. For scalar equations, this limit is determined by local data at the drift
equilibria and boundary points. We show that such localization fails for systems: switching between components can make some intervals spectrally relevant. We first prove that the logarithmic transforms of all component eigenfunctions converge to the same function, which solves a scalar Hamilton-Jacobi equation. The associated Aubry set determines the asymptotic behavior of the principal
eigenvalue. Under nondegeneracy assumptions, this set decomposes into isolated points, regular intervals, and composite intervals containing common drift zeros. We then identify the effective value of each class through local Ornstein-Uhlenbeck spectra and interval transport problems, and establish that the limit of the principal
eigenvalue is the minimum of these class values. A unique minimizing class also determines the common logarithmic eigenfunction profile.
Read the full abstract ▾
Artificial Intelligence
World Agent: Can Language Models Keep a World Running?
Weixing Chen, Weipeng Zhang, Nan An et al. (+2 more)
World models are moving from generating realistic frames to generating playable worlds, yet whether a delivered world can keep running is not tested anywhere. Existing evaluations stop at generation, at delivery, or at single-step transitions, and each stops at a different point along the way. Correct local state transitions or intermediate outcomes do not guarantee a correctly organized causal event flow. We propose the world
agent task, which moves the evaluation point of world generation from the moment of delivery to the continued operation that follows. In this task, a model is not asked to generate a world. It is held responsible for keeping the world running, which requires coordinating events and carrying forward their consequences to constrain subsequent
evolution. We instantiate the task in WorldAgent-
Benchmark with two complementary tracks. In the maintenance track, the model must ground the events of a continuous narrative into correct transitions of the explicit world state while respecting causal, temporal, and concurrency constraints. In the deduction track, the model must predict how the world will evolve under partial observations and act toward a goal. The maintenance track combines LLM-assisted semantic judgments with programmatic validation and scoring, while the deduction track is evaluated entirely programmatically. Individual judgments are auditable against world states and execution logs, and scores can be recomputed from the saved judgments and execution records. Across 8 models, scores decline steadily as pre-built structure is removed from the world, and causal-relation checking is the weakest component for every model. The
benchmark makes the continued operation of a world measurable and distinguishes local completion from failures in event organization. Code and
dataset will be released on https://github.com/HCPLab-SYSU/WorldAgent-
Benchmark.
Read the full abstract ▾
All authors Weixing Chen, Weipeng Zhang, Nan An, Yang Liu, Liang Lin
Cryptography and Security
Silent Failures in Agentic Security Evaluation: A Validated Harness for Tool-Call Mediation Under Indirect Prompt Injection
Animesh Shaw
LLM
agents that invoke privileged tools are vulnerable to indirect
prompt injection (IPI), in which
adversarial instructions embedded in retrieved data hijack the
agent's actions. A growing body of work evaluates defenses against IPI, but the validity of that evaluation is rarely examined. We audit an IPI
benchmark and its harness and identify four defect classes – silent payload non-delivery, attack success scored by tool identity rather than arguments, false-rejection rate conflated with model incapacity, and the absence of an audit trail – each of which yields a plausible, publishable, and incorrect number. We quantify the distortion by re-scoring identical execution traces under the defective and corrected definitions: on real
agent behaviour, the tool-identity scorer reports a 21.7% attack-success rate where the true argument-level rate is 1.2%. In the sharpest case, an open model previously reported at 62.8% registers 0% under the corrected harness – the prior figure largely an artifact of undelivered payloads and identity-level scoring. We release a harness whose construction makes each defect unrepresentable – machine-checkable payload placement, argument-level attacker predicates, per-scenario environments, and mandatory trace persistence – and use it to report three quantities the field does not: whether a compromised
agent discloses the attack, the full security/utility operating curve of an LLM-judge defense, and tool-calling capability disentangled from defensive over-blocking. A corrected harness further overturns a reported "capability barrier": a model deemed incapable of tool use is in fact fully capable, its earlier result an artifact of environment mismatch. We argue that evaluation validity is a prerequisite for, not a footnote to, defense claims in agentic security, and provide an instrument that enforces it.
Read the full abstract ▾
Computer Vision
MM-OPD: Towards One More Bottleneck Between Perception and Reasoning
Jintao Tong, Yujing Lou, Zhanming Shen et al. (+6 more)
Recent
multimodal large
language models (MLLMs) advance visual
reasoning by strengthening both perception and
reasoning, implicitly assuming a process that transitions seamlessly from perception to
reasoning. However, we observe a counterintuitive phenomenon that challenges this assumption: holding the model, question, and decoding fixed, we replace images with their caption or code representations (symbolic views), which seems to be redundant given the clear image structures, but the performance surprisingly improves by 10.2% to 23.6% across model scales and
datasets. We term this performance gap as the Symbolic Visual Gap and then take a closer look at it. Through experiments, we find that although the visual evidence can already appear in the
reasoning trace for the image-input model, the symbolic-view-input model shows much higher
attention to the correct evidence than the image-input model. This suggests that despite good capabilities from current works in perception and
reasoning themselves, another bottleneck exists between perception and
reasoning in selecting perceived visual information as appropriate evidence for subsequent
reasoning. To handle this bottleneck, since the symbolic view steers
attention toward correct evidence and is readily obtained at scale, it provides supervision for evidence selection without manually labeled evidence. Building on this, we introduce MM-OPD, a
multimodal on-policy self-
distillation framework for symbolic-to-visual correction that transfers guidance from symbolic-conditioned behavior to the image-conditioned policy through residual
token-level targets, steering the model toward correct visual evidence. Experiments across
benchmarks and model scales show that MM-OPD improves a broad range of
multimodal abilities, with gains in visual perception, chart and document understanding, mathematical
reasoning, and general VQA.
Read the full abstract ▾
All authors Jintao Tong, Yujing Lou, Zhanming Shen, Jiaqi Gu, Lubin Fan, Ruixuan Li, Yue Wu, Jieping Ye, Yixiong Zou
Machine Learning
Self-Evolving Time-Series Forecasting Agents with Episodic Memory and Online Policy Learning
Junyi Wang, Yilin Wang, Wen Wu et al. (+1 more)
LLM-based
agents are increasingly used for time-series
forecasting because they can organise contextual information, perform multi-step analysis, and guide the sequence of actions required to complete
forecasting tasks. Most existing
agents focus only on the current
forecasting instance. However, in real-world deployments,
forecasting commonly operates online, with new forecasts issued from the currently available history as the forecast origin advances and the ground-truth targets of earlier instances progressively become available. These targets provide feedback on the actions taken in earlier instances, yet existing
agents generally do not preserve or utilise this information to adapt their subsequent actions. To address this limitation, we introduce FASE, a Feedback-Aware Self-Evolving
forecasting agent that converts such feedback into task-specific experience for subsequent
forecasting instances. FASE combines episodic
memory, which retrieves relevant completed instances, with online policy learning, which summarises the feedback accumulated across instances into ranking guidance. The proposed framework is evaluated on 29
dataset configurations selected from the GIFT-Eval
benchmark. Across these 29 configurations, FASE attains the strongest aggregate point
forecasting performance among the evaluated methods and reduces the normalised MAE by 9.1% relative to the best individual foundation model baseline. The results further indicate that the cumulative advantage of FASE increases as delayed feedback accumulates. Together, these findings demonstrate that FASE can continually self-evolve through feedback from completed
forecasting instances without updating the parameters of the LLM.
Read the full abstract ▾
All authors Junyi Wang, Yilin Wang, Wen Wu, Chao Zhang
Combinatorics
The diameter of recoloring graphs under a maximum average degree bound
Ruilin Zheng and Junying Lu
For a graph $G$, we write $\mathrm{mad}(G)$ for its maximum average degree and $\mathrm{diam} G$ for its diameter. Let $R_k(G)$ be the graph whose vertices are the proper colorings of $G$ with $k$ colors, where two colorings are adjacent when they differ at one vertex. Feghali (JCTB, 2021) proved that, for fixed integers $d,k\ge 1$ with $k\ge d+1$ and every $\varepsilon>0$, every $n$-vertex graph $G$ satisfying $\mathrm{mad}(G)\le d-\varepsilon$ has $\mathrm{diam} R_k(G)=O_{d,k,\varepsilon}(n(\log n)^{d-1})$. In this article, we prove that \[ \mathrm{diam} R_k(G)=O_{d,k,\varepsilon}\!\left( n(\log n)^{\left\lfloor (d-1)/(k-d)\right\rfloor} \right), \] which extends the result proved by Feghali directly. The proof uses a partition into independent layers and removes $k-d$ colors at each recursive stage. We also improve the bound on the number of layers and determine the best possible linear coefficient in the forest case. More precisely, for $0<\varepsilon<2$, every $n$-vertex graph $G$ with $\mathrm{mad}(G)\le2-\varepsilon$ satisfies $\mathrm{diam} R_3(G)\le\rho_M n$, where $M=\lfloor2/\varepsilon\rfloor$, $\rho_M=\max_{1\le m\le M}D_m/m$, and $D_m$ is the largest diameter of $R_3(T)$ over all trees $T$ on $m$ vertices. Moreover, the coefficient $\rho_M$ is best possible.
Read the full abstract ▾
Artificial Intelligence
Dude, Where's My State? Execution Information Requirements for Stateful Agents
Nikita Mehrotra, Ashish Tiwari, Priyanshu Gupta et al. (+1 more)
Long-running
agents must preserve information that later steps depend on. We introduce the Execution Information Requirement (EIR), a lower bound on the information that must remain accessible for correct completion under specified task and access conditions. We develop LACUNA, a framework that generates tasks with known dependencies and varies information demand, retention, and recovery separately from the difficulty of individual operations. Across four models, restoring a missing result raises accuracy on affected recall steps to 100%, compared with 0% for equal-length irrelevant information. Sufficient storage alone does not ensure success: retention policies can discard required results, errors can propagate through later computations, and
agents can stop before recovery is complete. We also introduce VESTIGE, which uses
agent execution traces to construct semantic graphs and measure information demand for real tasks. Across 72,562 software-
agent trajectories, VESTIGE reveals a steeper distance-related decline in solution-relevant rereading for failed runs (RR 0.951 per distance doubling), while adjusted peak demand alone is not associated with failure. Together, these contributions support evaluating whether
agents preserve and recover the information their tasks require.
Read the full abstract ▾
All authors Nikita Mehrotra, Ashish Tiwari, Priyanshu Gupta, Sumit Gulwani
Optimization and Control
Optimal Reinsurance-Dividend Strategy with Fixed Transaction Costs in a Regime-Switching Brownian Risk Model: A Viscosity Solution to the Impulse Control Problem
Wenyuan Wang, Zuo Quan Xu and Kaixin Yan
We consider a problem of optimal proportional reinsurance-dividend distribution under a Brownian risk model, where both the drift and
volatility coefficients are subject to endogenous regime-switching. Dividend payments are subject to fixed transaction costs. The problem is formulated as a two-dimensional
stochastic control problem, and we prove that the value function is the unique viscosity solution of the associated Hamilton-Jacobi-Bellman equation with nonlocal operator. For almost all parameter configurations, we explicitly characterize the optimal strategy that maximizes the expected total discounted dividends net of transaction costs until ruin. The optimal dividend policy is a two-barrier impulsive strategy, while the optimal reinsurance proportion is given in feedback form. Numerical examples are provided to illustrate the optimality results.
Read the full abstract ▾
Artificial Intelligence
PINNMorph: Evolving Online Adaptation Policies for Physics-Informed Neural Networks
Xu Yang, Mingyang Yu, Jun Zhang et al. (+2 more)
Physics-informed neural networks (PINNs) provide a learning-based framework for solving partial differential equations (PDEs), yet their training behavior can change substantially throughout optimization. Residual distributions, gradient interactions, regional learning difficulty, and model-capacity requirements may evolve over time, while the network architecture and major training mechanisms are typically determined before training. We propose PINNMorph, an online PINN adaptation framework based on large
language model (LLM)-guided policy
evolution. PINNMorph maintains a population of state-conditioned adaptation policies that map execution diagnostics to controlled interventions over
topology modification, additive representation augmentation, objective balancing, gradient handling, adaptive sampling, and optimizer-phase control. At each intervention opportunity, candidate programs are instantiated from the current policy population, selected according to the observed training state, and applied directly to the PINN under training. The resulting model inherits its existing parameters and training state and continues optimization along the same trajectory. Execution outcomes are subsequently used to evaluate interventions and evolve the policy population. Unlike pre-training architecture search or fixed adaptation rules, PINNMorph jointly adapts the current PINN and the policies governing its interventions using feedback from actual training. Experiments on 13 PDE
benchmarks show that PINNMorph achieves lower solution errors than SA-PINN, ConFIG, RoPINN, HARMONIC, and PINNsAgent across all evaluated problems. Ablation studies further examine the effects of online adaptation, state-conditioned intervention selection, and execution-feedback-driven policy
evolution.
Read the full abstract ▾
All authors Xu Yang, Mingyang Yu, Jun Zhang, Keqian Li, Jing Xu
Computation and Language
Focusing Condition: Inference-Time Self-Contrastive Steering Elicits Better Conditional Text Embeddings in LLMs
Zifeng Cheng, Lingyun Qian, Zhiwei Jiang et al. (+5 more)
Extracting conditional text
embeddings from large
language models (LLMs) is a promising paradigm, as it requires neither additional data nor fine-tuning. Existing methods incorporate conditions into
prompts to guide LLMs to focus on specific aspects and elicit conditional text
embeddings. However, relying solely on
prompts often fails to produce high-quality conditional text
embeddings, as they remain entangled with general text
embeddings, ultimately degrading their quality. To this end, we propose an inference-time, plug-and-play Self-Contrastive Steering (SCS) method that constructs unconditional general text
embeddings and uses them to refine conditional text
embeddings, making them more focused on the target condition. Specifically, we modify the
attention mask and positional encodings to mask the condition, thereby obtaining unconditional text
embeddings and intervening in the multi-head self-
attention computation process. Notably, our method is highly efficient, requiring only a single additional multi-head self-
attention computation at inference time. Extensive experiments on clustering, Semantic Textual Similarity, and triplet
alignment datasets demonstrate that our method can seamlessly improve the performance of existing
prompt-based methods across different LLMs in a training-free and plug-and-play manner. Our code will be released at https://github.com/zifengcheng/SCS
Read the full abstract ▾
All authors Zifeng Cheng, Lingyun Qian, Zhiwei Jiang, Cong Wang, Yafeng Yin, Fei Shen, Ao Zhou, Qing Gu
Optics
Reconfigurable Linear Optical Transformations in a Single Integrated Multimode Waveguide
Manoah van der Worp, Jonas Krimmer, Ivo M. Vellekoop et al. (+2 more)
Wavefront shaping enables control over optical fields for applications ranging from imaging to photonic information processing. While conventional wavefront shaping relies on free-space systems comprising bulk optical components, compact and fully integrated approaches are comparatively unexplored. Here, we introduce a concept based on a single multimode waveguide with distributed thermo-optic perturbations for controlling multimode propagation. We develop a differentiable physical model and numerically demonstrate shaping the outgoing wavefront and programmable linear optical transformations, using gradient-based optimization. We achieve diffraction-limited focusing with 20 spatial modes and show that the theoretical focusing limit can be retained while maintaining more than $99\%$ of the input optical power. We extend this approach to complete input–output transformations and implement Sylvester–Hadamard
matrix operations with dimensions up to $32\times32$ with correlations exceeding $99\%$. Reducing the number of tunable elements requires stronger individual perturbations, resulting in stronger and broader mode coupling that extends beyond the target modes. To mitigate this effect, we consider ancillary modes and realize a $4\times4$ transformation that reaches $93.8\%\pm4.8\%$ correlation with only $5N^2$ tunable elements, where $N=4$ denotes the number of spatial modes. The best-performing realization reaches $98.3\%$ correlation with the target. Our results demonstrate that thermo-optic control of multimode waveguides offers an approach to fully-integrated wavefront shaping and programmable linear optical transformations, with potential applications in classical and
quantum photonic processing
Read the full abstract ▾
All authors Manoah van der Worp, Jonas Krimmer, Ivo M. Vellekoop, Pepijn W. H. Pinkse, Stefan Rothe
Signal Processing
Realistic modeling of Photonic Terahertz Communication based on a real-world 30 km 30 Gbps experimental demonstration
Mingxu Wang, Jianjun Yu, Xianming Zhao et al. (+9 more)
Experimental studies have demonstrated that THz links can support multi-kilometer, gigabit-per-second wireless transmission. However, few studies have effectively validated channel models through high-capacity, long-distance experiments. In this work, we experimentally demonstrate a 30 km ultra-long-haul photonic THz wireless communication system. 16 Gbaud quadrature phase shift keying (QPSK) signals are transmitted and tested, achieving a maximum data rate of 30 Gbps. For the first time, we validate two widely used THz channel models using a fully implemented 30 km experimental setup, providing a systematic and thorough analysis that integrates theoretical modeling with experimental results. We also perform an analysis of power consumption in photonic THz communication systems and provide a detailed comparison between simulated and experimental power consumption results. Finally, a detailed performance comparison between the simulation and experimental results, including signal-to-noise ratio (SNR), bit error rate (BER) and data rate, are provided. Our work provides essential insights for the design and energy-efficient deployment of future high-capacity THz networks, establishing a
benchmark for subsequent research in channel modeling, experimental verification, and system-level energy efficiency.
Read the full abstract ▾
All authors Mingxu Wang, Jianjun Yu, Xianming Zhao, Jiali Chen, Ye Zhou, Xin Lu, Hansong Ma, Chengzhen Bian, Wen Zhou, Kaihui Wang, Weiping Li, Iman Tavakkolnia
Computer Vision
RCVLA: 4D Radar-Grounded Semantic Reasoning and Trajectory Arbitration for Autonomous Driving
Lianqing Zheng, Xiaokai Bai, Yixuan Luo et al. (+6 more)
4D radar provides
geometric and motion cues that complement visual semantics, but integrating it into
vision-
language-action (VLA) models requires both radar–
language alignment for semantic
reasoning and explicit use of radar measurements for trajectory refinement and selection. To support these capabilities, we construct Cap4DR with 86,016 radar-image-text samples for
alignment pretraining and OmniHD-QA with 520,161 question-answer pairs for instruction tuning across scene description, key-object
reasoning, occupancy understanding, and trajectory planning. Building on these
datasets, we propose RCVLA, a radar-camera VLA framework consisting of a radar-grounded semantic
reasoning stage (RCVLA-Sem) and a trajectory arbitration stage (RCVLA-Phys). RCVLA-Sem performs gated bidirectional interaction between camera and radar
tokens for driving question answering and reference trajectory generation, while auxiliary heads provide object and occupancy queries. RCVLA-Phys refines reference-guided trajectory candidates through truncated
diffusion conditioned on these queries and cluster-level radar measurements, then calibrates candidate scores using radar-derived time-to-collision risk. On OmniHD-QA, RCVLA-Sem improves CIDEr by 9.92 points and reduces key-object velocity error by $21.9\%$ relative to OmniDrive. RCVLA-Phys further reduces average L2 error from $0.348$ to $0.259\,\mathrm{m}$ and average open-loop collision rate from $0.576\%$ to $0.175\%$ relative to RCVLA-Sem. Ablation studies further show that
language-aligned radar
tokens improve semantic
reasoning, while cluster-level radar measurements and risk calibration improve trajectory arbitration. Code will be released.
Read the full abstract ▾
All authors Lianqing Zheng, Xiaokai Bai, Yixuan Luo, Runwei Guan, Minghao Liu, Zhiqiang Wei, Hui-liang Shen, Xichan Zhu, Zhixiong Ma
Differential Geometry
Positive sectional curvature and non-isometric circle actions on a family of eleven-spheres
Shaoqiang Deng, Zhiguang Hu and Hui Zhang
We study the construction of positively curved Riemannian metrics by non-isometric circle actions. For a family of homotopy eleven-spheres whose Eells–Kuiper invariants form the even subgroup of $\mathbb{Z}/992$, we construct, on each member, a smooth background metric $q$ and three effective circle actions with generators $W_1,W_2,W_3$ such that the metric determined by $g^{-1}=q^{-1}+\sum_{a=1}^3W_a\otimes W_a$ has positive sectional
curvature. Each action is non-isometric for every partial metric, including its incoming metric and the final metric. We first describe the sphere by gauge transformations of the quaternionic
Hopf bundle. We then construct compatible metrics on two disks and smooth their inverse metrics while preserving the action formula. A local conjugation makes the circle actions non-isometric. For each fixed member, we obtain an explicit positive lower bound $2^{-54}(1+M_{0,k}+M_{1,k})^{-28}$, where $M_{0,k}$ and $M_{1,k}$ are norms of the
curvature and its first covariant derivative for its fixed connection. The bound may depend on the member of the family.
Read the full abstract ▾
Machine Learning
The GUI Is Not the State: Diagnosing State Aliasing in GUI World Models
Dongsheng Liu, Chao Jin, Wenkui Yang et al. (+7 more)
GUI World Models (GUI-WMs) are increasingly used to predict future states for
agent planning and simulation, yet most existing formulations condition only on the current GUI observation and action. We identify state aliasing, where the vis- ible interface omits transition-relevant environment state, so identical observable conditions can correspond to different valid futures. To diagnose this failure mode, we introduce StateAliasBench, a diagnostic
benchmark that explicitly isolates such ambiguities via strict pairing. We further propose lightweight predictive- state recovery that infers structured state from history and augments otherwise frozen GUI-WMs through a deterministic state interface. Family-specific special- ists provide state recovery across heterogeneous state types, and multi-teacher dis- tillation consolidates them into a single unified
estimator. Experiments show that existing GUI-WMs exhibit systematic failures under observation-only condition- ing, while predictive-state augmentation substantially restores state-sensitive pre- diction across evaluated WMs, preserves generative fidelity, and improves down- stream performance of GUI
agents on AndroidWorld. These results suggest that reliable GUI world modeling should account not only for what is visible, but also for the hidden transition state that determines what happens next.
Read the full abstract ▾
All authors Dongsheng Liu, Chao Jin, Wenkui Yang, Hejin Wang, Junwei Yang, Zeren Zhang, Ziwei Chen, Huaibo Huang, Jie Cao, Ran He
Statistics Theory
Reference-Null Calibrated Thresholds for E-Processes with Applications to Conformal Martingales
Yi Ding, Lan Wei, Xuehu Zhu et al. (+1 more)
E-processes provide a flexible framework for anytime-valid inference, with the conventional rejection boundary $1/\alpha$ typically justified by Ville's inequality. Such a boundary is universal but can be conservative, as it does not exploit additional information about the null distribution. We propose a reference-null calibration framework that uses independent null samples to construct sharper rejection thresholds while preserving type-I error control. To study the statistical gain from sharper thresholds, we specialize the framework to conformal martingales. Building on histogram-based betting, we incorporate Krichevsky–Trofimov smoothing and develop a restart-mixture construction for distribution shift detection. We establish quantitative results for detection power and detection delay and characterize how the reduction in the rejection boundary translates into power and delay gains relative to the conventional Ville boundary. Numerical experiments corroborate the theoretical findings across a range of distributional changes. Finally, we apply the proposed calibration strategy to an existing e-process for online LLM watermark detection, showing that the method can improve sequential detection performance without modifying the underlying e-process. These results demonstrate that reference-null calibration provides a general and modular way to enhance the efficiency of e-process-based sequential inference.
Read the full abstract ▾
All authors Yi Ding, Lan Wei, Xuehu Zhu, Wenlin Dai
Artificial Intelligence
When Better Gets Worse: Improvement Fidelity for Self-Improving Agents in Adaptive Worlds
Ke Wang, Zijie Zhao, Zhiyi Yuan et al. (+1 more)
Self-improving
agents increasingly rely on proxy verifiers to choose policy updates, yet deployment can change the world in which those updates are evaluated. An update that looks better to the verifier can therefore become worse after deployment even when the verifier ranks policies well overall. We formalize this gap as Improvement Fidelity, which asks whether proxy improvements preserve the sign and ordering of deployment improvements over the updates an improvement process actually proposes. We show that global policy accuracy need not guarantee update fidelity: operator shift and deployment response can create update-level errors, while candidate margins determine whether those errors change the replacement decision. We introduce PIVOT-KG, a paired, decision-aware validator that allocates scarce high-fidelity evaluation according to the expected reduction in selection regret per unit cost. Across 90 held-out roots in Leduc, Kuhn, and Melting Pot, proxy and deployment optimal sets are disjoint in 51 cases. In an eight-candidate HighwayEnv stress test, PIVOT-KG reduces mean improvement-selection regret from 0.0435 under the exact Uniform validation rule to 0.0055 at the primary budget. Together, these results show why reliable self-improvement should evaluate proposed improvements in the worlds they induce, while providing a practical rule for allocating scarce deployment evidence when it can affect the replacement decision.
Read the full abstract ▾
All authors Ke Wang, Zijie Zhao, Zhiyi Yuan, Changlun Li
Machine Learning
SIFT: Enhancing Time Series Foundation Models via Semantic Invariance and Structural Fidelity Fine-Tuning
Yi Tang, Tengxue Zhang, Yang Shu et al. (+3 more)
Time Series Foundation Models (TSFMs) have achieved remarkable zero-shot performance through extensive pre-training on massive time series
datasets. Nevertheless, due to the low-dimensional properties and diverse structural patterns of time series data, performing naive fine-tuning on TSFMs often leads to overfitting and falling into the mean-prediction trap. To address these challenges, we propose SIFT, a robust adaptation method that enhances time series foundation models by preserving Semantic Invariance and structural Fidelity throughout the fine-Tuning process. We employ semantic-invariant
adversarial augmentation, which utilizes semantic spectrum decomposition to partition the semantic space and then generates perturbations within the non-core semantic subspace to bolster the model's robustness against these perturbations, mitigating overfitting. We implement a component-based structural fidelity enhancement, which facilitates component-wise mixup and imposes a reconstruction objective to improve the model's ability to preserve structural fidelity, alleviating the mean-prediction trap. Extensive experiments on representative TSFMs covering 10 real-world
datasets demonstrate that SIFT can significantly enhance the performance of TSFMs.
Read the full abstract ▾
All authors Yi Tang, Tengxue Zhang, Yang Shu, Chenjuan Guo, Chenchen Sun, Yisheng An