arXiv subjects

Tao Yang

Publications and source records attributed to Tao Yang.

At least 19 recordsLinked to original sources

arXiv2026

Visually Irrelevant Hallucinations in Medical Vision-Language Models

Medical vision-language models (VLMs) show strong performance in radiology report generation (RRG) and medical visual question answering (Med-VQA), producing free-form clinical descriptions and answers from medical images. However, they may generate clinically meaningful statements that lack visual evidence or contradict image content. We refer to these errors as visually irrelevant hallucinations and investigate them at the phrase level. We propose CLEAR (Counterfactual Localized Evaluation of Alignment and Response), a model-agnostic post-hoc framework for phrase-level hallucination risk estimation. CLEAR decomposes generated responses into visually assessable clinical phrases. For each phrase, phrase-conditioned grounding identifies a candidate visual region, which is selectively perturbed by a localized counterfactual intervention. CLEAR characterizes the intervention using two complementary auditing cues: response cues, capturing semantic and clinically meaningful changes in the generated response, and alignment cues, measuring changes in image-phrase compatibility. These cues are normalized and jointly modeled to estimate phrase-level hallucination risk. Across four datasets spanning RRG and Med-VQA, CLEAR outperforms state-of-the-art hallucination detection baselines across multiple medical and general-purpose VLMs. On MIMIC-CXR, CLEAR improves AUPRC from 85.10% to 90.23%. Its effectiveness is further validated on a GT-derived controlled VIH benchmark. These results establish visually irrelevant hallucination as an important fine-grained reliability failure mode in medical VLMs and demonstrate the value of model-agnostic visual-support auditing for more trustworthy multimodal clinical generation.

cs.CV↗

arXiv2026

CoinRAG: Contextualized Information Nugget KV Cache Reuse for Long-Context RAG

Recent optimization studies on Retrieval-Augmented Generation (RAG) have exploited chunk-level KV cache reuse to avoid processing long retrieved contexts for higher efficiency, while significant information redundancy and noise still remain in the coarse-grained chunks. This paper optimizes the Pareto frontier under low prefill latency constraints while maximizing accuracy by proposing CoinRAG (Contextualized Information Nugget KV Cache Reuse for Long-Context RAG). The name metaphorically reflects our core mechanism: much like assembling small tokens (or "coins") to accumulate a larger value, CoinRAG compositionally reuses offline-computed, fine-grained nugget caches to form a learned contextual representation efficiently in a more semantically relevant but compact manner. Specifically, instead of full-chunk encoding, CoinRAG identifies query-relevant semantic units within retrieved chunks through two-stage retrieval and seamlessly assembles their sliced KV representations with a chunk-level context. CoinRAG can be integrated with CacheBlend's selective recomputation for further accuracy gains under the same latency budget. Extensive evaluations on LongBench multi-hop question answering tasks demonstrate that CoinRAG substantially shortens the active context and outperforms the other baselines with a new Pareto frontier and an average 5.3% relative improvement in answer quality (F1) under a standard fast prefill latency budget.

cs.CL↗

arXiv2026

Ultra-Fast 4H-SiC AC-LGAD for 4D Tracking Applications

4H-SiC low gain avalanche detectors (LGADs) combine internal gain with the high carrier saturation drift velocity, thermal stability, and low leakage current of SiC. AC-coupled readout (AC-LGAD) adds position information to the precise timing, which enables 4D tracking. This work reports the first experimental demonstration of a 4H-SiC AC-LGAD. Devices with a $1.0\times1.0$~mm$^2$ active area and a $2\times2$ array of AC-coupled pads at a pitch of 500~$μ$m were fabricated with two gain-layer doping concentrations, $4\times10^{17}$ and $5\times10^{17}$~cm$^{-3}$, and characterized with a 375~nm ultraviolet transient current technique (UV-TCT) scanning setup. The hit position and time were reconstructed with the conventional discretized positioning circuit (DPC) and constant fraction discrimination (CFD) methods. They were also reconstructed with a transformer-based neural network, which processes the full waveforms of the four pads through a shared encoder and formulates timing as a per-sample segmentation task. With DPC and CFD, the best position and time resolutions are about 10~$μ$m and 20~ps. The neural network mitigates the nonlinear distortions near the device edges and improves the position resolution to below 5~$μ$m and the time resolution to about 15~ps, improvements of more than 60\% and 30\%, respectively. The higher-doped device reaches a high signal-to-noise ratio at lower bias and keeps good performance at low laser intensities. These results demonstrate the potential of 4H-SiC AC-LGADs for 4D tracking in future collider experiments.

physics.ins-det↗

arXiv2026

EndoVLA-OOV: Language-Conditioned Out-of-View Target Acquisition in Robotic Endoscopy

The scan-then-revisit workflow of upper gastrointestinal endoscopy requires bringing a previously observed suspicious region back into view for closer inspection. Acquiring such targets requires translating anatomical and appearance descriptions into camera motions from partial observations of a confined, deformable environment. We present EndoVLA-OOV, a language-conditioned vision-language-action policy for out-of-view target acquisition in robotic endoscopy. The policy adapts a pretrained backbone to continuous two-degree-of-freedom distal bending commands and combines initial, intermediate, and current images to guide search. To support policy learning, we provide an open-source VLA dataset containing 1,050 episodes across six tasks and approximately 262k observation-action pairs. Physical evaluation spans wet-lens and low-light conditions, two unseen stomach phantoms, and cross-embodiment transfer to an unseen ex-vivo environment. In the cross-embodiment evaluation, the full-data policy achieves 93.3% target acquisition success (28/30). In a controlled training-data comparison, expanding the corpus from 312 to 1,050 episodes increases acquisition success from 40.0% to 90.0% under the same model, training, and test configuration. Instruction and observation interventions further support the use of language and temporal context in the evaluated acquisition tasks. These results support language-guided endoscopic revisitation across the evaluated environments and robot embodiments and demonstrate the benefit of expanded task-specific demonstrations.

cs.RO↗

arXiv2026

Dynamic Regret and Tightened Constraint Violation for Communication-Efficient Distributed Online Bilevel Optimization

This paper investigates the distributed online bilevel optimization problem with time-varying inequality constraints. In this setting, a multi-agent network is confronted with a hierarchical decision-making task characterized by coupling between the inner and outer optimization levels. We first propose a communication-efficient single-loop distributed online primal--dual algorithm based on communication compression, which tracks both the inner-level solution and the Hessian-inverse-vector product using only one update per-round. We show that this algorithm achieves a dynamic network regret bound of $\mathcal{O}(T^{\max\{θ_1,1-θ_1\}}+T^{θ_1 } P_T+H_{1,T}+E_{1,T})$, where~$θ_1 \in (0,1)$, $P_T$ is the outer-level path variation, $H_{1,T}$ is the inner-level path variation, and $E_{1,T}$ is the gradient path variation alongside a network cumulative constraint violation bound of $\mathcal{O}(T^{1-θ_1/2})$, which is further reduced to $\mathcal{O}(T^{1-θ_1})$ under Slater's condition. Furthermore, to eliminate the dependence of regret on the inner-level and gradient path variation, we propose a novel adaptive multi-loop distributed online algorithm to control the inner-level approximation accuracy at each round. By dynamically adjusting the number of inner-level updates based on a gradient-norm tolerance condition, this algorithm yields an improved dynamic network regret bound of $\mathcal{O}(T^{\max\{θ_1,1-θ_1\}}+T^{θ_1 } P_T)$ which depends solely on the outer-level path variation while maintaining the same network cumulative constraint violation bound. Finally, numerical simulations verify the effectiveness of the proposed algorithms.

math.OC↗

arXiv2026

GUMPIM: Unitary and Malleable Memory for Processing-in-Memory with Guaranteed PIM Pages

DRAM-based Processing-in-Memory (PIM) addresses the "memory wall" by executing computations directly inside main memory. However, memory interleaving and virtual memory limit contiguous data size visible to PIM units, constraining PIM task granularity. To mitigate this, existing PIM systems drastically isolate PIM memory or disable memory interleaving. These design choices, however, decrease the CPU memory bandwidth and introduce extra data transfer, leading to an additional "system memory wall" that must be resolved to realize PIM's full potential. In this work, we propose GUMPIM, a PIM system that allows interleaved CPU pages and non-interleaved PIM pages to coexist in a Unitary and Malleable memory space with Guaranteed PIM page allocation. GUMPIM enables zero-copy during PIM task offloading and maintains CPU memory bandwidth while ensuring low PIM offloading overhead. First, we propose a dual-track memory management mechanism consisting of independent page allocation and address translation for CPU and PIM pages. Second, we design GUMPIM interface hardware on PIM-enabled DRAMs to provide a dynamic address mapping for the different data layouts of CPU and PIM pages. Third, we propose a PIM-assisted page migration mechanism that transparently migrates pages while preserving CPU access bandwidth, thereby enabling guaranteed and accelerated PIM page allocation. GUMPIM requires no changes to commodity DRAM standards; all hardware modifications are limited to the DRAM side, ensuring full compatibility with existing CPUs and enabling immediate deployment on current HBMx- and LPDDRx-based PIM platforms. Our results show only <0.1% performance degradation for CPU workloads on GUMPIM. For PIM workloads, GUMPIM reduces memory allocation and CPU-part computation times by 2.7x and 4.93x, respectively, yielding an end-to-end 2.3x speedup over a state-of-the-art baseline system.

cs.AR↗

arXiv2026

Mock Catalogs of Strongly Lensed Gravitational Waves via a Halo Model Approach with Space-borne Detectors

Future space-borne gravitational-wave (GW) detectors, such as LISA and DECIGO, are expected to detect a large number of GW events, a fraction of which may be strongly lensed by intervening galaxies or galaxy clusters. In this work, we develop a comprehensive framework to simulate strongly lensed GWs in the context of space-borne detectors. Based on realistic astrophysical models for both the source population and the lens distribution, we construct mock catalogs of lensed GW events, referred to as \textbf{GW-LMC-Space}. Our results show that, for a four-year LISA observation, the expected number of lensed events ranges from $0$ to $131$, depending on the adopted formation model of massive black hole binaries (MBHBs). The corresponding lensing probability for MBHBs can reach up to $\sim 0.3\%$. For DECIGO, we find that the number of lensed events in a one-year observation is expected to lie in the range of $0$--$44$, with a lensing probability of $\sim 0.15\%$ for stellar-mass binary black holes (BBHs), binary neutron stars (BNSs), and neutron star--black hole binaries (NSBHs). We further show that the overlap of lensed signals is a common feature in space-borne detectors, which can significantly affect both the signal-to-noise ratio (SNR) estimation and event identification. These results highlight the importance of accounting for signal overlap in the analysis of strongly lensed GW events in future space-borne GW observations.

astro-ph.CO↗

arXiv2026

Process Advantage Signal Shaping: Advantage Construction for Process-Supervised Reinforcement Learning in LLM Reasoners

Dense process supervision supplies local feedback but leaves open how that feedback should become a policy advantage. We introduce PASS (Process Advantage Signal Shaping), an advantage-construction pipeline for group-relative policy optimization (GRPO). PASS combines separate reward-channel normalization and fusion, signal-defined chunks for credit accumulation, and normalization by the remaining chunk count. This design connects the scale of local feedback to the resolution and horizon over which it contributes to a policy update. We evaluate PASS using learned process rewards for mathematical reasoning and teacher-derived token signals for multi-hop question answering. On seven mathematics benchmarks, PASS improves average pass@1 by $4.62$ points over outcome-only GRPO while reducing average response length by approximately $2\%$. On multi-hop QA with mean-centered normalization, PASS exceeds outcome-only training by $0.87$ and $0.63$ points for two distillation signals, with longer responses. Across both normalization schemes, PASS achieves the highest mean accuracy among the tested process-signal configurations and remains closest to outcome-only under absolute-maximum normalization. These findings show that channel handling, credit resolution, and accumulation should be considered jointly when incorporating dense supervision into policy optimization.

cs.AI↗

arXiv2026

Integer-Forcing Precoding: A Geometric Perspective and Low-Complexity Algorithm Design

The joint coefficient--scaling design of integer-forcing (IF) precoding couples a full-rank integer coefficient matrix with a positive diagonal scaling matrix, resulting in a mixed discrete--continuous optimization problem. We first distinguish the original full-rank formulation and its finite-field invertibility requirement from the unimodular restriction studied in the subsequent analysis. Under this restriction, we formulate coefficient selection as the unimodular minimum-sum basis problem (UMSBP). By parameterizing the scaling matrix with its normalized diagonal vector, we define a deterministic exact single-output UMSBP oracle whose preimages partition the scaling manifold into disjoint coefficient-labelled regions. The restricted objective is consequently represented as the lower envelope of fixed-coefficient quadratic functions. Based on this formulation, we derive an exact decomposition into fixed-coefficient scaling optimization, coefficient-consistency verification, cross-region exploration, and candidate comparison. We prove that exact UMSBP evaluation is NP-hard through a polynomial-time reduction from exact Euclidean CVP. For a fixed coefficient matrix, we derive an equivalent convex formulation, its normalized stationarity condition, and a matrix-balancing interpretation. Under the stated positivity and finite-projective-diameter condition

cs.IT↗

arXiv2026

Does Target Alignment Mean Target Recovery? An Evidence-Ladder Study of Adversarial Claims on Contrastive Encoders

Adversarial attacks on vision-language models optimize an image toward a text target, then cite the attacked model's similarity score as evidence of success. We ask whether that score - victim-space target alignment (VTS) - predicts recovery of the target by an independent model. We first build a measurement instrument: supervised judges outside the attacked geometry, real-target blend controls, shuffled-target negatives, and a reference level derived from a 50% target-image blend. Two preregistered studies then compare six contrastive encoders under a matched attack at three perturbation budgets. Robustly trained encoders (FARE, TeCoA, PMG, TRADES) transfer substantially more independent evidence than vanilla CLIP or SigLIP; all eight contrasts reject at the bootstrap floor. However, no cell reaches the blend-derived reference level. The three best cells fall within its replication band, leaving practical recovery undecided. Within robust encoders, per-sample alignment gain correlates with evidence gain ($ρ= 0.24-0.51$); within vanilla CLIP the correlation is consistent with zero. Across encoders we find no monotone alignment-evidence relation. VTS is therefore informative only within a fixed robust encoder, and we provide a reporting protocol in its place.

cs.CV↗

arXiv2026

Evidence for orbital eccentricity supports a hierarchical origin for GW231123

GW231123 is the most massive binary black hole merger observed to date and one of the most intriguing hierarchical merger candidates in the pair-instability mass gap. Its exceptionally large inferred component spins and remarkably short in-band duration make the source highly sensitive to waveform physics near merger. Here we show that GW231123 is better described as an eccentric binary black-hole merger. We analyse data over 11--448 Hz using the eccentric, spin-precessing effective-one-body waveform model TEOBResumS--Dalí, accounting for detector calibration uncertainty. We measure $e_{5.5\,\mathrm{Hz}}=0.34^{+0.12}_{-0.23}$ at 90\% credibility and obtain a Bayes factor of 10.2 in favour of the eccentric model over its quasi-circular counterpart. GW231123-like injection--recovery studies further show that analyses starting at 20 Hz, as in previous studies, can obscure this eccentricity. Including eccentricity substantially changes the spin inference, weakening the preference for a near-extremal secondary spin and lowering its posterior median to $a_2\simeq0.7$, close to the characteristic spin expected for a black-hole merger remnant. These results provide a coherent dynamical interpretation of GW231123, in which residual eccentricity, a mass-gap black hole, and a remnant-like spin arise naturally from a hierarchical merger channel in a dense environment.

astro-ph.HE↗

arXiv2026

Adaptive Sparsity Optimization with Learnable Soft Top-K and Per-Term Thresholding for Efficient Retrieval

Recent work on neural sparse retrieval has demonstrated strong relevance by leveraging Large Language Models (LLMs) for semantic term expansion. However, learned models paired with previous sparsification techniques still yield overly long document and query vectors partly due to a large LLM vocabulary, imposing a serious challenge to retrieval time and space efficiency. This paper proposes a scheme for optimizing model sparsity through a synergy of adaptive strategies, including learnable soft top-K, per-term thresholding, and FLOPs regularization to increase the sparsity of query and document vectors. Experimental results with Lion-SP model on the MS MARCO and BEIR datasets demonstrate that the proposed scheme can outperform the baselines by significantly reducing the average query and document lengths. Our scheme can achieve much shorter retrieval latency and lower storage cost while maintaining highly competitive relevance.

cs.IR↗

arXiv2026

Impact of Gain Layer Doping Concentration on the Performance of 4H-SiC LGADs

Gain layer doping concentration determines the electric field profile in a low gain avalanche detector (LGAD) and is therefore the primary design parameter. This work reports its influence on the charge collection and timing behavior of 4H-SiC LGADs. Two mesa-isolated devices with cold He ion implanted termination differ only in gain layer doping, $4\times10^{17}$~$cm^{-3}$ and $5\times10^{17}$~$cm^{-3}$, and are characterized together with a PIN diode from the same wafer using the ultraviolet transient current technique (UV-TCT). The 25\% doping difference produces a gain difference of about two orders of magnitude at the same applied bias, representing a gain range of 3.7 to 216. Under UV-TCT excitation, the higher-doped device exhibits time-resolution minima within a gain range of about 20--80 across the investigated laser intensities, consistent with the transition from electronic-noise-dominated jitter to multiplication shot noise. The non-monotonic timing response is also observed under a $^{90}Sr$ beta source with a Si LGAD reference: the higher-doped device reaches 51.7 ps at 280 V and then degrades, while the lower-doped device requires more than 450 V. This measured gain window provides an experimental reference for further optimization of the gain-layer design.

physics.ins-det↗

arXiv2026

DRAM: Delta-rule Recurrent Associative Memory for Robot Manipulation Policies

Robotic manipulation is inherently history-dependent, yet most pretrained robotic policies condition on only the current observation or a short temporal window. Equipping such policies with long-term memory remains challenging: existing approaches either feed the backbone multi-frame observation windows, which substantially increase inference cost, or rely on pre-defined semantic features, which limit task generality and may also require the retraining of the backbone to adapt to the memory. We introduce DRAM (Delta-rule Recurrent Associative Memory), a plug-and-play memory module that can be attached to a wide range of pretrained robotic policies, endowing them with long-horizon memory without architectural modification or backbone retraining, requiring only task-specific post-training of the memory module and action expert. DRAM maintains a fixed-size associative memory using gated delta-rule linear attention, with a modified update that incorporates all tokens within each frame in parallel. An architecture-agnostic readout integrates historical context into action prediction across different policy architectures. Experiments show that DRAM consistently improves frozen pretrained policies over short-context baselines and alternative compact memory designs, validating its effectiveness as a fixed-size, post-hoc memory module trained with the backbone frozen.

cs.RO↗

arXiv2026

Probabilistic Symbolic-Distillation Model of Droplet Collision for Spray Simulation at High Ambient Pressures

Droplet collision governs droplet population dynamics in many chemical engineering processes, such as spray drying, spray cooling, agricultural spraying, and combustion. Existing analytical models impose deterministic, pairwise boundaries between collision outcomes, whereas machine-learning classifiers lack the explicit functional form required of analytical collision submodels. In this study, we develop a probabilistic symbolic-distillation model using nearly forty thousand experimental events spanning eight regimes and five dimensionless parameters, including over five thousand data for ambient pressure up to 50 atm. A machine-learning teacher learns the joint outcome-probability landscape from these data, and symbolic regression subsequently distils it into eight class-specific expressions that jointly define a coupled analytical model. The resulting analytical field replaces abrupt regime switching with finite-width fuzzy boundaries. It outperforms the evaluated conventional analytical boundary models and reveals that their main limitation is the inability of zero-width boundaries to represent gradual probability transitions. The "biased-dice" sampling scheme provides a statistically consistent and practically convenient model implementation for Eulerian-Lagrangian spray simulation.

physics.flu-dyn↗

arXiv2026

Certificate-Carrying Distributed Model Predictive Control on Product Manifolds with $\mathrm{SO}(3)$

This paper studies constraint certification in synchronous distributed model predictive control (DMPC) when neighboring predictions change between sampling instants. Before the parallel local solves, each agent communicates a shifted prediction and an announced update budget. A hard trajectory trust region makes that budget enforceable, while an edge-wise feasibility cap computed from the shifted packets keeps the fallback feasible without using any current optimizer output. Distance and relative-attitude constraints are tightened with explicit Lipschitz constants and two budget layers: one accounts for the simultaneous neighbor update and the other retains a checkable shift reserve. We prove hard pairwise constraint satisfaction and recursive feasibility under stated nominal-execution and terminal assumptions, give the additional residual caused by execution error, and derive a local practical value-decrease bound. A spacecraft formation example uses hard terminal and pairwise constraints, a geodesic relative- attitude constraint on $\SO$, and reproducible terminal-set checks. Comparisons with fixed, trajectory-only, and windowed online margins show that the proposed budget reduces conservatism while preserving a positive shifted-feasibility margin.

eess.SY↗

arXiv2026

Evaluation of OpenAI o1: Opportunities and Challenges of AGI

This comprehensive study evaluates the performance of OpenAI's o1-preview large language model across a diverse array of complex reasoning tasks, spanning multiple domains, including computer science, mathematics, natural sciences, medicine, linguistics, and social sciences. Through rigorous testing, o1-preview demonstrated remarkable capabilities, often achieving human-level or superior performance in areas ranging from coding challenges to scientific reasoning and from language processing to creative problem-solving. Key findings include: -83.3% success rate in solving complex competitive programming problems, surpassing many human experts. -Superior ability in generating coherent and accurate radiology reports, outperforming other evaluated models. -100% accuracy in high school-level mathematical reasoning tasks, providing detailed step-by-step solutions. -Advanced natural language inference capabilities across general and specialized domains like medicine. -Impressive performance in chip design tasks, outperforming specialized models in areas such as EDA script generation and bug analysis. -Remarkable proficiency in anthropology and geology, demonstrating deep understanding and reasoning in these specialized fields. -Strong capabilities in quantitative investing. O1 has comprehensive financial knowledge and statistical modeling skills. -Effective performance in social media analysis, including sentiment analysis and emotion recognition. The model excelled particularly in tasks requiring intricate reasoning and knowledge integration across various fields. While some limitations were observed, including occasional errors on simpler problems and challenges with certain highly specialized concepts, the overall results indicate significant progress towards artificial general intelligence.

cs.CL↗

arXiv2026

Hunyuan-A13B Technical Report

We present Hunyuan-A13B, an open-source large language model based on a Mixture-of-Experts architecture. It contains 80 billion total parameters but activates only 13 billion during inference, balancing model capability, computational efficiency, and deployment cost. The model is pretrained on a rigorously filtered 20T-token corpus with enhanced STEM data curation, improving factual reliability and reasoning ability. High-quality supervised fine-tuning and large-scale reinforcement learning further enhance its overall performance. Hunyuan-A13B also introduces a dual-mode Chain-of-Thought framework that adapts reasoning depth to task complexity: fast thinking for routine queries and slow thinking for complex, multi-step problems. Evaluations show competitive performance across mathematics, science, programming, general language understanding, and agent tasks, often approaching that of much larger models. Its high inference throughput makes it suitable for latency-sensitive applications. We release Hunyuan-A13B to support open research and practical LLM deployment.

cs.AI↗