Searcharxiv⌕ Search

arXiv subjects

Chengxi Li

Publications and source records attributed to Chengxi Li.

At least 19 recordsLinked to original sources
arXiv2026

DiabAgent: Toward Agentic Physiological Control

As pretraining scales, medical language models have advanced rapidly in clinical question answering. Their ability to achieve effective physiological control through sustained intervention still lacks systematic quantitative evaluation. Type 1 diabetes provides a concrete setting: impaired insulin secretion creates dependence on exogenous insulin, with dosing continually adjusted using information about meals, exercise, illness, and glucose measurements. We introduce DiabAgent, a research suite for agentic physiological control comprising the Agentic Physiological Control (APC) framework, the DiabetSim simulator, and the SweetBench benchmark. APC controls glucose through medical-device tools, using ongoing observations to infer individual states, accumulate personal experience, and choose treatment actions. We build DiabetSim to simulate different patients, events, and device conditions for scalable virtual testing and training. Using DiabetSim, we construct SweetBench by instantiating 32 scenario designs as 212 glucose-control tasks. With Claude Opus 5.5 and a fixed decision interval of one hour, APC achieves a COGI of 73.5, compared with 71.2 for the open-source closed-loop baseline OpenAPS. Ablations further show that individual adaptation and contextual information improve glucose control. Combining high-frequency rule-based control with lower-frequency agent decisions shows potential for balancing control performance and computational cost. Our work provides a framework, simulator, and evaluation foundation for agents to participate in human physiological control.

cs.AI↗
arXiv2026

EPIC: Rethinking Trust Regions for Autoregressive LLM Policies via Exact-KL Prefix Importance-Ratio Clipping

Reinforcement learning (RL) based on PPO-style clipping has become an effective approach to improving the reasoning capabilities of large language models (LLMs). However, standard token-level clipping measures policy change only through the conditional probability of the current token, while the prefix conditioning context is itself generated by the policy. This mismatch becomes especially consequential in long-response and off-policy training. We reformulate autoregressive generation as a finite-horizon Markov decision process and derive its performance-difference identity under behavior-policy sampling. The resulting change of measure identifies the cumulative likelihood ratio of the generated prefix as the exact importance weight for each stepwise policy-improvement term. Starting from the monotonic-improvement principle of trust-region policy optimization, we then construct a prefix-level trust region using the exact reverse-KL coordinate $r-1-\log r$, whose behavior-policy expectation equals the corresponding prefix-distribution KL divergence. This yields EPIC, a PPO-style sequence-clipping method that directly regulates prefix-level policy drift without relying on a local quadratic approximation or expert-routing replay. Experiments on off-policy reinforcement learning with a sparse mixture-of-experts language model show that EPIC consistently outperforms standard and sequence-level clipping baselines across mathematical-reasoning benchmarks, while remaining competitive with routing-replay-based alternatives. Code is available at https://github.com/LiChengxi666/EPIC.

cs.LG↗
arXiv2026

Cooperative Boundary Switching of Mesoscopic Real Spectra in a Non-Hermitian Chain Family

At fixed bulk criticality, every fixed nonsingular real single-end potential preserves a growing local population of real eigenvalues in an open analytic family of non-Hermitian chains. An open region of two-end real potentials removes this population. A boundary-amplitude criterion fixes the minimum number of controlled ends. The population grows as $\sqrt N$, where $N$ is chain length, although its normalized spectral weight vanishes and the local limiting bands remain unchanged. For a representative chain's four zero-detuning settings, an independent passive Hermitian double chain converts this distinction into separated resonances at one fixed local port under a common coupling scan at fixed probe carrier. Its absolute reflection contrast and stability margin both scale as $N^{-1}$, with explicit sufficient resource budgets. Boundary amplitudes thus fix a minimum control requirement that limiting bands and thermodynamic spectral density do not specify.

cond-mat.mes-hall↗
arXiv2026

The noise cost of reservoir locality in broadband quantum amplifier chains

Directional quantum amplifiers enable weak-signal readout while protecting the source, but reservoir noise limits sensitivity. Finite-band suppression must reconcile frequency-dependent responses with fixed local couplings and finite rates. We construct adjacent-site loss channels attaining the known single-frequency minimum under local compatibility conditions, as an uncapped benchmark. For specified uniform and dimerized chains, we compare fixed reservoirs at matched drift, ports, high-gain band and complete gain/loss rate caps. At the representative cap $R=8$ and halfwidth $Ω=0.05$, nearest-neighbor arithmetic band average noise exceeds the unrestricted optimum by more than $1.01265$ quanta in the uniform chain for every finite $N\ge64$. Boundary-response bounds give necessary rates and targets requiring range two. At $N=64$, finite-auxiliary calculations retain a noise advantage under matched total coupling budgets; response bounds quantify noise-resolution tolerances. These theoretical results guide reservoir design and noise calibration for a proposed cryogenic microwave array with a $100$ kHz design band.

quant-ph↗
arXiv2026

ChronoScribe: Semantic-Numerical Specialization for Instruction-Driven Time Series Forecasting

Recent advances in instruction-driven time series forecasting highlight a fundamental tension: systems must simultaneously master natural-language instruction following and precise numerical prediction. Current approaches often force a compromise. Large language models (LLMs) can interpret unstructured requests but may be inaccurate when asked to produce forecast values directly, whereas time series foundation models (TSFMs) provide numerical forecasts but typically require structured inputs specifying the target series, context window, and horizon. We present ChronoScribe, which applies semantic-numerical specialization to this workflow. A LoRA-adapted LLM proposes an execution path and context length; deterministic rules resolve the executable configuration. On the specialist path, a frozen TSFM produces quantile trajectories for the task-specified horizon. Deterministic trajectory summaries condition the response, while the LLM neither generates nor revises predicted values. We evaluate a Qwen3.5-9B--Chronos-2 implementation across 23 evaluation dimensions and 30 public datasets against eleven proprietary direct LLMs and three open TSFMs. ChronoScribe has the lowest error on sMAPE, MASE, WAPE, RMSE, and MAE among all evaluated systems, including fixed-context controls. Across these metrics, even the lowest direct-LLM error remains 2.0--2.7 times the corresponding ChronoScribe error. At the system level, ChronoScribe achieves a Task-Level Score (TLS) of 65.5, 10.6 points above the strongest direct-LLM baseline. The results support ChronoScribe's division of labor, in which the LLM handles interpretation, configuration, and communication while the TSFM handles numerical forecasting.

cs.AI↗
arXiv2026

Characterizations for quasiconformal weighted Bergman spaces

We establish characterizations of planar quasiconformal mappings in weighted Bergman spaces $A_α^p(\mathbb D)$. For $p>0$ and $α>-1$, our characterization involves the convergence of certain integrals involving superlevel set areas, the Hausdorff measure of boundary superlevel sets, and the averaged derivative. These include quasiconformal analogues of characterizations established by Karafyllia and Karamanlis for conformal mappings. Moreover, for $-1<α\leq 0$, we obtain an additional equivalent characterization in terms of the Hausdorff measure of level sets. The power $α$ is sharp: for every $α>0$, we construct a bounded $K_α$-quasiconformal mapping for which the weighted level-set length integral with exponent $α+2$ diverges.

math.CV↗
arXiv2026

Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition

Adversarial games have driven advances from heuristic search to reinforcement learning, yet learning and adapting strategies from limited samples remain challenging. AI agents offer an alternative by turning game experience into revisions of executable policies. Building on heuristic learning (HL), we formalize Adversarial Heuristic Learning (AHL), a paradigm that uses AI agents as learning engines to refine game policies and supporting software while keeping model weights fixed. We introduce AAArena, a benchmark comprising 12 authentic adversarial games and 1,920 archived human programs, with an evaluation protocol modeled on real-world game competitions. Agents interpret rules, choose opponents, analyze replays, and revise game agents to achieve their highest ranking within fixed match and evaluation budgets. We evaluate \val{completedmodels} model and harness configurations: Opus5.5 with Claude Code earns 6 gold medals, while no evaluated configuration tops the remaining 6 human ladders. Performance is generally weaker in games with more complex rule specifications. Further experiments show that opponent selection and dense feedback support policy improvement, and that agents learn from both on-policy replays of their own matches and off-policy replays of other players' matches. These results highlight HL's potential in adversarial games and identify persistent challenges in game understanding, strategy implementation, and long-horizon policy development.

cs.AI↗
arXiv2026

VE2VF: Vision-Enabled to Vision-Free Distillation via Real-world Reinforcement Learning for Robust Contact-Rich Manipulation

When using reinforcement learning (RL) for contact-rich robotic manipulation, vision can provide task-relevant information that complements robot proprioception. However, vision-enabled policies tend to overfit to the visual conditions seen during training, limiting their robustness and transferability. We present a human-in-the-loop RL framework that employs teacher-student distillation to achieve robust performance across multiple task variants, trained entirely in the real world without requiring domain randomization or data augmentation. A vision-enabled teacher distills its knowledge into a vision-free student that relies solely on pose, twist, and wrench sensing, combining fast training with strong task generalization. On the real-world NIST assembly benchmark board, our approach achieves 95% overall success after approximately 50 minutes of training on 3 representative tasks, including robust generalization to 8 unseen task variants. Fine-tuning with distillation achieves full success on the most challenging task. We demonstrate that the resulting policies outperform baselines in both robustness and adaptability. Page: https://tuwien-asl.github.io/VE2VF/.

cs.RO↗
arXiv2026

DualManip: Agentic Dynamic Manipulation via Dual-Path Semantic Reasoning and Geometric Adaptation

Vision-language models (VLMs) enable open-vocabulary reasoning for robot manipulation, but their high inference latency limits responsiveness in dynamic scenes. Many scene changes, however, alter object geometry without invalidating task intent. We present DualManip, a dual-path framework that decouples infrequent semantic reasoning from responsive geometric adaptation. The semantic path decomposes the task and grounds task-relevant interactions, followed by a constraint-solving module for pose optimization. During execution, the geometric path continuously updates template-to-observation correspondences from live RGB-D observations via a shape-adaptive network. These correspondences transfer task-relevant grasp contacts across observations, enabling online grasp reconstruction under object motion and non-rigid deformation. The Information Interaction Module bridges the two paths by initializing task-relevant grasps from semantic grounding, validating geometric updates, and triggering semantic replanning upon update failures. Real-world evaluation spans six manipulation tasks covering non-rigid deformation, articulated reconfiguration, rigid motion, and high-precision assembly across three settings: static, single-change, and continuous dynamic. DualManip demonstrates superior manipulation robustness, particularly under continuous scene changes, while achieving geometric adaptation approximately 46$\times$ faster than agentic verification and semantic replanning. Our project page: https://lichengxi1.github.io/Dualmanip.

cs.RO↗
arXiv2026

Efficient and Adaptive Simultaneous Speech Translation with Fully Unidirectional Architecture

Simultaneous speech translation (SimulST) produces translations incrementally while processing partial speech input. Although large language models (LLMs) have shown strong capabilities in offline translation tasks, applying them to SimulST poses notable challenges. Existing LLM-based SimulST approaches either incur significant computational overhead due to repeated encoding of bidirectional speech encoder, or they depend on a fixed read/write policy, limiting the efficiency and performance. In this work, we introduce Efficient and Adaptive Simultaneous Speech Translation (EASiST) with fully unidirectional architecture, including both speech encoder and LLM. EASiST includes a multi-latency data curation strategy to generate semantically aligned SimulST training samples and redefines SimulST as an interleaved generation task with explicit read/write tokens. To facilitate adaptive inference, we incorporate a lightweight policy head that dynamically predicts read/write actions. Additionally, we employ a multi-stage training strategy to align speech-text modalities and optimize both translation and policy behavior. Experiments on both in-domain (MuST-C) and out-of-domain (Europarl-ST) En-De and En-Es datasets demonstrate that EASiST offers superior latency-quality trade-offs compared to several strong baselines.

cs.CL↗
arXiv2026

Surfsvr: 2D Surface Priors as 3D Geometric Regularizers for Sparse Voxel Reconstruction

Sparse voxel reconstruction offers an efficient representation for high-fidelity 3D modeling, yet its geometry is commonly optimized from local photometric evidence and discrete visibility statistics. This often leads to fragmented surfaces, excessive subdivision, and floating artifacts, particularly in weakly textured or sparsely observed regions. We introduce SurfSVR, a novel sparse voxel reconstruction paradigm that treats 2D surface priors as explicit 3D geometric regularizers. Instead of directly lifting noisy pixel-wise depth predictions, SurfSVR first organizes each image into coherent surface regions by jointly reasoning over appearance, monocular depth, normals and cross-view geometry. Each region is then represented by an adaptively selected planar or quadratic surface model based on fitting reliability and geometric complexity, while cross-model agreement distinguishes reliable geometry from ambiguous predictions. These structured 2D priors are lifted into 3D and integrated throughout the reconstruction pipeline. They guide surface-adaptive voxel subdivision, provide region-level depth and normal supervision during optimization, enhance geometrically reliable sparse-observed surfaces in voxel pruning, and suppress off-surface floaters during post-refinement training. This unified design converts semantic and geometric coherence in image space into persistent structural constraints in 3D. Extensive experiments on 3 public benchmarks demonstrate that SurfSVR consistently improves sparse voxel reconstruction across scenes with substantially different visibility and geometry characteristics, achieving state-of-the-art reconstruction quality. Codes and models will be released soon.

cs.CV↗
arXiv2026

Uncertainty-Aware Cross-Modal Remote Sensing Image-Text Retrieval via Evidential Learning

In cross-modal remote sensing image-text retrieval (CMRSITR), test-time remote sensing (RS) images and textual descriptions may deviate from well-curated benchmark conditions due to sensor- and atmosphere-related image degradations and text-side RS-vocabulary heterogeneity. Under such non-ideal conditions, existing CMRSITR methods may produce unreliable retrieval results because they perform retrieval with full certainty for each query and do not distinguish the varying uncertainty across queries. To address this issue, we propose an evidential learning-based CMRSITR (ELC) method for uncertainty-aware retrieval. During the training phase of ELC, evidential learning (EDL) is employed to model the inter-modal correspondences between RS images and textual descriptions as Dirichlet distributions, from which the uncertainty of each query can be obtained. Based on the EDL outputs, uncertainty-correctness alignment learning (UCL) is introduced to align the estimated uncertainty with retrieval correctness, encouraging high uncertainty for incorrect retrieval and low uncertainty for correct retrieval. Furthermore, intra-modal relationship learning (RL) distills the intra-modal similarity structure from pretrained mentor encoders for the trainable encoders, thereby making the Dirichlet distributions modeled by EDL more discriminative. In the test phase of ELC, the estimated uncertainty is compared with a threshold determined by a fixed deferral ratio, where low-uncertainty queries are directly returned and high-uncertainty queries are refined by RS-aware test-time augmentation (RS-TTA). Experimental results demonstrate that ELC achieves competitive retrieval performance compared with state-of-the-art CMRSITR methods and provides stronger robustness under the evaluated RS-specific degradations, including sensor- and atmosphere-related image perturbations and RS-vocabulary heterogeneity.

cs.IR↗
arXiv2026

Uniformization of cofat domains on metric two-spheres

We extend \emph{Schramm's cofat uniformization theorem} to cofat domains on upper Ahlfors 2-regular metric two-spheres $X$. Specifically, we show that if $Ω\subset X$ is a cofat domain, then there exists a $\fracπ{2}$-quasiconformal homeomorphism $f: Ω\to D$ onto a circle domain $D \subset \mathbb{S}^2$. Moreover, $f$ preserves the point-components and non-trivial complementary components. We also construct examples which show that the above conclusions are not true for countably connected $\ell^α$-subdomains of $\mathbb{S}^2$.

math.CV↗
arXiv2026

DexTwist: Dexterous Hand Retargeting for Twist Motion via Mixed Reality-based Teleoperation

Dexterous teleoperation via Mixed Reality (MR)-based interfaces offers a scalable paradigm for transferring human manipulation skills to dexterous robot hands. However, conventional retargeting approaches that minimize kinematic dissimilarity (e.g., joint angle or fingertip position error) often fail in contact-rich rotational manipulation, such as cap opening, key turning, and bolt screwing. This failure stems from the embodiment gap: mismatched link lengths, joint axes/limits, and fingertip geometry can cause direct pose imitation to induce tangential fingertip sliding rather than stable object rotation, resulting in screw axis drift, contact slip, and grasp instability. To address this, we propose DexTwist, a functional twist-retargeting framework for MR-based dexterous teleoperation. DexTwist detects a tripod pinch, estimates the operator's intended screw axis and twist magnitude, and applies a real-time residual joint-space refinement that tracks turning progress while regularizing the robot tripod geometry. The refinement minimizes a virtual-object objective defined by turning angle, screw axis consistency, fingertip closure, and tripod stability. Simulation and real-world experiments show that DexTwist improves turning angle tracking and screw axis stability compared with a vector-based retargeting baseline.

cs.RO↗
arXiv2026

Coded Robust Aggregation for Distributed Learning under Byzantine Attacks

In this paper, we investigate the problem of distributed learning (DL) in the presence of Byzantine attacks. For this problem, various robust bounded aggregation (RBA) rules have been proposed at the central server to mitigate the impact of Byzantine attacks. However, current DL methods apply RBA rules for the local gradients from the honest devices and the disruptive information from Byzantine devices, and the learning performance degrades significantly when the local gradients of different devices vary considerably from each other. To overcome this limitation, we propose a new DL method to cope with Byzantine attacks based on coded robust aggregation (CRA-DL). Before training begins, the training data are allocated to the devices redundantly. During training, in each iteration, the honest devices transmit coded gradients to the server computed from the allocated training data, and the server then aggregates the information received from both honest and Byzantine devices using RBA rules. In this way, the global gradient can be approximately recovered at the server to update the global model. Compared with current DL methods applying RBA rules, the improvement of CRA-DL is attributed to the fact that the coded gradients sent by the honest devices are closer to each other. This closeness enhances the robustness of the aggregation against Byzantine attacks, since Byzantine messages tend to be significantly different from those of honest devices in this case. We theoretically analyze the convergence performance of CRA-DL. Finally, we present numerical results to verify the superiority of the proposed method over existing baselines, showing its enhanced learning performance under Byzantine attacks.

cs.LG↗
arXiv2026

Biased Compression in Gradient Coding for Distributed Learning

Communication bottlenecks and the presence of stragglers pose significant challenges in distributed learning (DL). To deal with these challenges, recent advances leverage unbiased compression functions and gradient coding. However, the significant benefits of biased compression remain largely unexplored. To close this gap, we propose Compressed Gradient Coding with Error Feedback (COCO-EF), a novel DL method that combines gradient coding with biased compression to mitigate straggler effects and reduce communication costs. In each iteration, non-straggler devices encode local gradients from redundantly allocated training data, incorporate prior compression errors, and compress the results using biased compression functions before transmission. The server aggregates these compressed messages from the non-stragglers to approximate the global gradient for model updates. We provide rigorous theoretical convergence guarantees for COCO-EF and validate its superior learning performance over baseline methods through empirical evaluations. As far as we know, we are among the first to rigorously demonstrate that biased compression has substantial benefits in DL, when gradient coding is employed to cope with stragglers.

cs.DC↗
arXiv2026

Byzantine-Robust and Communication-Efficient Distributed Training: Compressive and Cyclic Gradient Coding

In this paper, we study the problem of distributed training (DT) under Byzantine attacks with communication constraints. While prior work has developed various robust aggregation rules at the server to enhance robustness to Byzantine attacks, the existing methods suffer from a critical limitation in that the solution error does not diminish when the local gradients sent by different devices vary considerably, as a result of data heterogeneity among the subsets held by different devices. To overcome this limitation, we propose a novel DT method, cyclic gradient coding-based DT (LAD). In LAD, the server allocates the entire training dataset to the devices before training begins. In each iteration, it assigns computational tasks redundantly to the devices using cyclic gradient coding. Each honest device then computes local gradients on a fixed number of data subsets and encodes the local gradients before transmitting to the server. The server aggregates the coded vectors from the honest devices and the potentially incorrect messages from Byzantine devices using a robust aggregation rule. Leveraging the redundancy of computation across devices, the convergence performance of LAD is analytically characterized, demonstrating improved robustness against Byzantine attacks and significantly lower solution error. Furthermore, we extend LAD to a communication-efficient variant, compressive and cyclic gradient coding-based DT (Com-LAD), which further reduces communication overhead under constrained settings. Numerical results validate the effectiveness of the proposed methods in enhancing both Byzantine resilience and communication efficiency.

cs.DC↗
arXiv2026

LLaDA2.1: Speeding Up Text Diffusion via Token Editing

While LLaDA2.0 showcased the scaling potential of 100B-level block-diffusion models and their inherent parallelization, the delicate equilibrium between decoding speed and generation quality has remained an elusive frontier. Today, we unveil LLaDA2.1, a paradigm shift designed to transcend this trade-off. By seamlessly weaving Token-to-Token (T2T) editing into the conventional Mask-to-Token (M2T) scheme, we introduce a joint, configurable threshold-decoding scheme. This structural innovation gives rise to two distinct personas: the Speedy Mode (S Mode), which audaciously lowers the M2T threshold to bypass traditional constraints while relying on T2T to refine the output; and the Quality Mode (Q Mode), which leans into conservative thresholds to secure superior benchmark performances with manageable efficiency degrade. Furthering this evolution, underpinned by an expansive context window, we implement the first large-scale Reinforcement Learning (RL) framework specifically tailored for dLLMs, anchored by specialized techniques for stable gradient estimation. This alignment not only sharpens reasoning precision but also elevates instruction-following fidelity, bridging the chasm between diffusion dynamics and complex human intent. We culminate this work by releasing LLaDA2.1-Mini (16B) and LLaDA2.1-Flash (100B). Across 33 rigorous benchmarks, LLaDA2.1 delivers strong task performance and lightning-fast decoding speed. Despite its 100B volume, on coding tasks it attains an astounding 892 TPS on HumanEval+, 801 TPS on BigCodeBench, and 663 TPS on LiveCodeBench.

cs.LG↗