Search arXiv⌕ Search

arXiv subjects

Xiang Li

Publications and source records attributed to Xiang Li.

At least 55 records · Page 3Linked to original sources
arXiv2026

Beyond State-Wise Mirror Descent: Offline Policy Optimization with Parametric Policies

We investigate the theoretical aspects of offline reinforcement learning (RL) under general function approximation. While prior works (e.g., Xie et al., 2021) have established the theoretical foundations of learning a good policy from offline data via pessimism, existing algorithms that are computationally tractable (often in an oracle-efficient sense), such as PSPI, only apply to finite and small action spaces. Moreover, these algorithms rely on state-wise mirror descent and require actors to be implicitly induced from the critic functions, failing to accommodate standalone policy parameterization which is ubiquitous in practice. In this work, we address these limitations and extend the theoretical guarantees to parameterized policy classes over large or continuous action spaces. When extending mirror descent to parameterized policies, we identify contextual coupling as the core difficulty, and show how connecting mirror descent to natural policy gradient leads to novel analyses, guarantees, and algorithmic insights, including a surprising unification between offline RL and imitation learning.

cs.LG↗
arXiv2026

CoVA: Composing VAE Representations for Generative Learning

Representation alignment with pretrained visual encoders improves diffusion transformer training but requires auxiliary encoders and extra training overhead. Recent approaches instead exploit representations from the generative model's own variational autoencoder (VAE). However, reconstruction-oriented VAE features are not optimized for the semantic discrimination and spatial structure needed for generative learning. We introduce \textbf{CoVA} (\textbf{Co}mposed \textbf{VA}E Representations), a lightweight framework that composes VAE features online into stage-wise guidance for flow-matching training. A small latent composer is jointly trained with class supervision, masked latent reconstruction, and horizontal-flip consistency, yielding semantically discriminative, spatially structured dense features and sample-dependent class posteriors. CoVA aligns intermediate SiT features with the composed representations early in training, then removes the increasingly detrimental alignment and reuses detached composer posteriors for sample-dependent soft label conditioning. We further uncover a hidden confound in soft label conditioning: averaging class embeddings attenuates their magnitude, coupling semantic interpolation with conditioning strength. This attenuation alone can substantially improve FID while degrading Recall, revealing a non-semantic source of the apparent gain. CoVA corrects this confound by restoring the hard-label magnitude, isolating posterior-dependent directional semantics. CoVA adds only 8.5M training-only auxiliary parameters (1.3\% of SiT-XL/2) and introduces no inference-time overhead. On ImageNet 256$\times$256, CoVA reduces SiT-XL/2 FID from 17.2 to 7.4 (\textbf{57\%}) in 400K steps. With classifier-free guidance, it achieves an FID of \textbf{2.03} in only 400 epochs, versus 2.06 after 1,400 epochs for vanilla SiT-XL/2.

cs.CV↗
arXiv2026

S-Space: Exploring Spatial Workspace in Multimodal Language Models

Multimodal language models (MLLMs) exhibit promising spatial reasoning capabilities, yet the internal mechanisms supporting them remain poorly understood. We uncover S-Space, a linear representation subspace in which object-token representations encode continuous coordinates along horizontal, vertical, and distance axes. Although we identify these axes by differentiating spatial-answer scores with respect to object-token activations, S-Space is intrinsic rather than induced by specific data or prompts, remaining readable under non-spatial questions. We show that S-Space functions as an internal spatial workspace of MLLMs. Perception encodes object locations into it, forming spatial representations that can be directly read out to support accurate spatial judgments. Internal reasoning transforms it for changing viewpoints. Verbal responses selectively access it, evidenced by causal interventions that alter spatial judgments while largely preserving non-spatial responses. Case studies further show that S-Space assigns locations to unseen entities and the observer, and connects physical directions with abstract linguistic meanings, suggesting both the capabilities and limitations of large-scale language training. Finally, we use S-Space as a diagnostic tool for spatial reasoning failures. Direction-word readouts reveal systematic coupling between cardinal and image-relative reference frames. More importantly, a perception-computation decomposition exposes a key bottleneck: MLLMs can form capable spatial representations, yet fall short in the spatial transformations required to reason reliably across viewpoint changes.

cs.CV↗
arXiv2026

CoMMa: Contribution-Aware Medical Multi-Agents for Decentralized Oncology Decision Support

Recent multi-agent frameworks have shown promise for oncology decision support, yet most assume centralized data access and rely on prompt-based assignment, limiting their applicability in privacy-sensitive clinical settings. We propose Contribution-Aware Medical Multi-Agents (CoMMa), a decentralized LLM-agent framework where specialists operate on partitioned clinical data streams. Unlike prior approaches that share inputs across agents, CoMMa enforces data decentralization to include stronger role specialization and further enhances this via agent-specific finetuning. To enable reliable and interpretable coordination, we introduce a contribution-aware aggregation mechanism that replaces stochastic, narrative-based reasoning with deterministic embedding projections to approximate each agent's marginal utility. This yields explicit credit assignment over agents, providing a stable and interpretable decision pathway aligned with clinical requirements. We evaluate CoMMa on multiple oncology benchmarks, including real-world multidisciplinary tumor board datasets from large academic hospitals in North America and East Asia, as well as public datasets, demonstrating strong performance and generalization across heterogeneous clinical settings.

cs.AI↗
arXiv2026

One Frame, Full Heartbeat: ECG-Free Cardiac Cine MRI Synthesis via Phase-Conditioned Flow Matching

Cine cardiovascular magnetic resonance (CMR) analysis relies on multi-frame sequences capturing the full cardiac cycle. However, standard multi-frame acquisition depends heavily on electrocardiogram (ECG) gating and repeated breath-holds, posing challenges in uncooperative populations, resource-limited settings, and temporally corrupted datasets. Existing methods that synthesize full cardiac sequences either rely on explicit ECG signals to parameterize myocardium function, or employ deformable registration without physiological constraints, failing to faithfully reproduce clinically relevant dynamic metrics such as ejection fraction (EF) and ventricular contraction magnitude. We present PhaseFlow, a unified generative framework that overcomes both limitations. PhaseFlow estimates a non-linear cardiac phase signal directly from the input sequence via a segmentation-derived left-ventricular (LV) area curve, capturing the asymmetric dynamics of systole and diastole without any ECG dependency. At inference, this phase signal is provided by a pathology-specific template, informing phase-specific frame generation. A rectified flow model conditioned on the phase and slice position synthesizes the full cardiac motion trajectory in the latent space, decoded into a diffeomorphic displacement field that warps end-diastole pixel intensities directly, eliminating the reconstruction blur often accompanying the variational autoencoder. On the ACDC benchmark, PhaseFlow achieves superior physiological fidelity and image realism, with best LV volume curve $R^2$, structural similarity (SSIM) and generative quality (FID) among all baselines. Ablation studies confirm that each proposed component contributes measurably to the overall performance.

cs.CV↗
arXiv2026

A GPU-Parallel Framework for Heterogeneous Multi-Task Reinforcement Learning

GPU-parallel simulation provides abundant robot interaction, but existing benchmarks rarely combine this scale with heterogeneous manipulation tasks and standardized multi-task RL evaluation. We introduce Hebero (Heterogeneous Benchmark for Robot Learning), a GPU-parallel Isaac Lab benchmark that enables efficient joint training and evaluation of a single policy across all 40 heterogeneous tasks. Scaling experiments show that increasing parallel replicas per task improves success under a fixed wall-clock budget. To support learning with sparse rewards and limited demonstrations, we propose Demonstration-Guided Policy Optimization (DGPO), which reuses demonstrations for dense tracking rewards and asymmetric value learning. Its shared stack supports controlled comparisons of learner-specific demonstration interfaces within PPO. Within DGPO framework, we introduce IW-ABC, which uses a lightweight per-task learning progress signal to coordinate adaptive behavior cloning (ABC), relaxing demonstration guidance with task progress, and importance weighting (IW), emphasizing lagging tasks in PPO updates. With 50 demonstrations per task, IW-ABC achieves 90.1% state-input mean success, outperforming the strongest baseline FAMO-ABC by 7.8 percentage points. Its visual counterpart reaches 93.5% mean success. Real-world experiments further demonstrate that a single multi-task policy trained in simulation can successfully perform four tasks on a physical Piper robot. The project page is available at https://hebero-rl.github.io/.

cs.RO↗
arXiv2026

Stationary common-neighborhood properties and partition hypotheses

We use stationary common-neighborhood properties to study highly connected Ramsey relations and partition hypotheses. For weakly compact $κ$, $\operatorname{Coll}(ω_1,{<}κ)$ forces $ω_2\to_{\mathrm{hc},<5}(ω_2)^2_ω$ and $\operatorname{PH}_1(ω_2)$. If $κ$ is $T^{κ^+}_{ω_1}$-Ramsey, the same collapse forces that every countable coloring of $[ω_2]^2$ has a stationary set $X\subseteqω_2$ and a color $i$ such that every finite subset of $X$ has stationarily many color-$i$ common neighbors in $X$. From one weakly compact cardinal, we obtain a model of the ${<}5$-edge relation at $ω_3$ and $\operatorname{PH}_1(ω_3)$, in which $\check H^2(ω_3,A_d)\ne0$ for every nontrivial abelian group $A$. This separates $\operatorname{PH}_1(ω_3)$ from $\operatorname{PH}_2(ω_3)$, with the exact consistency strength of one weakly compact cardinal. We also show that $\operatorname{PH}_1(ω_2\timesω_5)$ is equiconsistent with two weakly compact cardinals.

math.LO↗
arXiv2026

QUARTET: Quad-branch cross-Attention and Random-walk Traces for Enhancing Transformers on Relational Graphs

Relational Deep Learning (RDL) models multi-table databases as heterogeneous temporal graphs, and graph transformers currently achieve state-of-the-art performance on benchmarks like RelBench. However, the current leading model, RelGT, suffers from two key limitations: its random local sampler yields loosely connected subgraphs that hinder message passing, and its global attention module relies on a single, seed-feature-based memory that ignores broader macro-level dynamics. To overcome these limitations, we introduce QUARTET, an expressive graph transformer architecture that applies full self-attention on local subgraphs while enriching global context through cross-attention branches. Specifically, QUARTET employs a Causal Random Walk (CRW) sampler based on recency-truncated Personalized PageRank (PPR) to extract compact, hub-robust, and densely connected local subgraphs without temporal leakage. Concurrently, a quad-branch cross-attention module integrates global context from four complementary perspectives: seed feature, seed topology, temporal dynamics, and collaborative dynamics. Across the RelBench v1 classification tasks, QUARTET consistently matches or outperforms the current state-of-the-art graph transformer baselines (HGT and RelGT). Ablation studies confirm that the CRW sampler significantly enriches local neighborhood quality, while the global branches provide essential, task-specific predictive gains.

cs.LG↗
arXiv2026

One Frame, Full Heartbeat: ECG-Free 4D Cardiac Cine MRI Synthesis via Radial-Decomposed Flow Matching

Cine cardiovascular magnetic resonance (CMR) captures the cardiac cycle as a four-dimensional (4D) sequence, but standard acquisition requires electrocardiogram (ECG) gating and repeated breath holds. Visual realism alone does not establish accurate patient-specific ejection fraction (EF) or ventricular volumes. We present PhaseFlow3D, a generative framework that synthesizes a complete 4D cine sequence from a single end-diastolic (ED) three-dimensional (3D) volume without ECG. To capture asymmetric systolic and diastolic dynamics, it represents the cardiac cycle as a piecewise linear phase anchored at ED and end-systolic (ES) time points. At inference, a population-level canonical template supplies this phase without patient-specific temporal information. A phase-conditioned rectified flow model generates a cardiac motion trajectory in latent space. Radial Contraction Decomposition converts each latent state into a 3D displacement field, combining a physics-informed radial component for centripetal myocardial contraction with an image-conditioned residual for rotation and out-of-plane motion. Each frame is generated by directly warping the ED volume, bypassing variational autoencoder decoding. On the combined ACDC and M&Ms benchmark, PhaseFlow3D achieves the lowest EF mean absolute error, the only positive left-ventricular volume-curve $R^2$, and the best distributional quality among compared methods. Ablations confirm each component's contribution. Downstream evaluations demonstrate the utility of the synthesized sequences and displacement fields for segmentation, pathology classification, label propagation, and myocardial strain analysis.

cs.CV↗
arXiv2026

A superconducting qutrit link beyond the qubit limit

Superconducting microwave links have enabled deterministic state transfer and remote entanglement between qubits, but deterministic links have so far operated with an effectively two-dimensional transmitted Hilbert space. Here we demonstrate a superconducting qutrit link between two independently packaged nodes connected by a microwave channel. Each node combines a transmon qutrit, a transmission resonator, and a tunable Purcell-filter interface, allowing the two remote microwave-photon interfaces to be matched in both frequency and bandwidth. We implement two transition-selective photon-mediated operations that transfer the $|e\rangle$ and $|f\rangle$ qutrit components in distinct temporal modes of the same channel. We tomographically characterize arbitrary qutrit-state transfer, obtaining a mean transferred-state fidelity of 83.68% and a qutrit process fidelity of 77.12%, exceeding both the classical qutrit-transfer benchmark and the best possible average fidelity of an effective qubit channel used to transmit an arbitrary qutrit. Using partial-transfer operations, we reconstruct a remote two-qutrit state with negativity 0.730, a tomography-inferred dense-coding capacity of 2.273 bits, and a tomography-inferred Collins-Gisin-Linden-Massar-Popescu (CGLMP) parameter $I_3=2.332$, all beyond the corresponding qubit or local bounds. These results demonstrate a superconducting microwave link that uses the native three-level structure of transmons as a genuine high-dimensional communication resource.

quant-ph↗
arXiv2026

Conformal Prediction with Paraphrase-Aware Scoring for LLM Uncertainty Quantification

Uncertainty quantification (UQ) for large language models (LLMs) aims to provide reliable measures of predictive confidence, yet current methods are often unstable under meaning-preserving perturbations. Semantically equivalent paraphrases can induce substantial variability in predictive confidence, even for methods with formal guarantees, such as conformal prediction. To address this issue, we propose a paraphrase-aware UQ framework robust to semantic rewordings. Our approach trains a lightweight proxy model on LLM hidden states and aggregates its predictions across paraphrases to construct label-wise nonconformity scores. Under score exchangeability, conformal calibration retains marginal coverage. This guarantee can also hold under test-only rewording, provided that the paraphrase pipeline satisfies an additional distributional alignment condition. We evaluate three settings (normal, fully reworded, and semi-reworded) which apply rewording to neither dataset, both calibration and test datasets, or only the test dataset, respectively. Across seven multiple-choice QA benchmarks and multiple model families, our method produces compact prediction sets with empirical coverage generally near the nominal target, even in the semi-reworded setting. Ablation studies show that the learned proxy accounts for most of the reduction in set size, while paraphrase-augmented training and inference-time aggregation improve stability under rewording. Code is available at https://github.com/Raina-Xin/PA_Score.

cs.CL↗
arXiv2026

When Talk Isn't Code: Comparing LLM Agents That Simulate and Develop Software

LLMs are used both to simulate network protocol implementations, as honeypots do, and to write them. Prior work evaluates the two uses separately, from a model's answers or conversations in one case and from its generated code in the other. We observe that a knowledge probe (asking the model which security checks an implementation needs) and a conversation can credit security checks that the generated program lacks, but no study has compared them with the code the same model writes. To fill this gap, we present the first such comparison between simulation mode (S-mode), where the model plays the implementation in a conversation, and development mode (D-mode), where the model writes the program and the program is attacked, with a knowledge probe as a baseline. We design and implement a harness that judges all three with one attack suite and one oracle, and evaluate 15 LLMs on four protocol implementations. We find that with the full security specification, the median model's attack success is at most 4\% on reassembly, HTTP, and firewall implementations in both modes, but 13\% in S-mode and 15\% in D-mode on DNS. In D-mode, adding the missing DNS requirement to the specification cuts that attack from 100\% to 34\% and changes only that check, whereas deleting a stated rule weakens the program on other checks too. Asking the model beforehand does not predict which checks a program has, since 13 of the 14 models that write a DNS resolver name the check, yet all 42 programs lack it. S-mode is a useful first filter, flagging 82\% of the real DNS vulnerabilities and leaving 96\% of the safe cases unflagged. It errs in both directions, over-reporting security checks that D-mode does not implement and under-reporting checks that D-mode does. S-mode can screen but not replace D-mode testing, and security requirements should be written down even when a model can recite them.

cs.SE↗
arXiv2026

Development of Small-pitch, Ultra-thin 3D Silicon Sensors at USTC

We report on the development of 3D silicon sensors at the University of Science and Technology of China (USTC). The sensor involves columnar electrodes (5 um in diameter) of both doping types, etched from the same wafer side. The p+ electrodes pass through the epitaxial wafer, whereas the n+ electrodes stop at a short distance from the opposite side of the epitaxial wafer. Based on the current processing capacity, the sensors feature an ultra-thin active substrate (50 um) and a small pixel size of 50 um x 50 um or 25 um x 25 um. This R&D project aims to establish a sensor technology to simultaneously measure position and time information at the single-pixel level. The first fabrication run, using a merged layout containing both pixel sizes, has been completed. The design and fabrication of the sensors, followed by their characterization, including Current-Voltage (I-V) measurements and Sr-90 beta source measurements, are reported in this paper.

physics.ins-det↗
arXiv2026

SWE-chat: Coding Agent Interactions From Real Users in the Wild

AI coding agents are being adopted at scale, yet we lack empirical evidence on how people actually use them and how much of their output is useful in practice. We present SWE-chat, the first large-scale dataset of real coding agent sessions collected from open-source developers in the wild. The dataset currently contains almost 18,000 sessions, comprising more than 229,000 user prompts and 2 million agent tool calls. SWE-chat is a living dataset; our collection pipeline automatically and continually discovers and processes sessions from public repositories. Leveraging SWE-chat, we provide an initial empirical characterization of real-world coding agent usage and failure modes. We find that coding patterns are bimodal: in 41% of sessions, agents author virtually all committed code ("vibe coding"), while in 25%, humans write all code themselves. Despite rapidly improving capabilities, coding agents remain inefficient in natural settings. Only 59% of all agent-produced code survives into user commits, and agent-written code introduces more security vulnerabilities than code authored by humans. Furthermore, users push back against agent outputs - through corrections, failure reports, and interruptions - in 50% of all turns. By capturing complete interaction traces with human vs. agent code authorship attribution, SWE-chat provides an empirical foundation for moving beyond curated benchmarks towards an evidence-based understanding of how AI agents perform in real developer workflows.

cs.AI↗
arXiv2026

GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution

The executable harness surrounding a GUI model determines how observations are assembled, actions are executed, and verification, recovery, and termination are controlled. Compared with harness optimization for non-GUI agents, automatically optimizing this harness poses three coupled challenges: reconciling model intent with observed visual effects, diagnosing failures under variable execution outcomes, and identifying recurrent failure patterns across tasks and translating them into reusable runtime changes. We introduce GUI-HARVEST, an automatic harness optimizer that enables self-improving GUI agents with frozen backbone models. First, to ground diagnosis in observed action effects, it aligns model outputs and executed actions with before-and-after screenshots, tying findings to specific interface transitions. Second, to account for execution variability, it treats repeated runs of the same task as a joint evidence unit, using within-task comparisons to locate outcome-relevant behavioral differences. Third, it consolidates verified findings across tasks into recurring failure patterns, maps them to bounded source-code edits with predictions recorded before evaluation, and checks the predicted behavioral effects alongside task performance through repeated execution. Experiments on OSWorld-Verified show consistent held-out gains across six general-purpose open, GUI-specialized open, and proprietary backbone models; Qwen3-VL-32B-Instruct gains 12.33 points on the full suite. Frozen-harness transfer improves GPT-5 by 13.87 percentage points on WindowsAgentArena at 50 steps without further optimization. With the same backbone and initial harness, GUI-HARVEST outperforms Self-Harness and Meta-Harness, suggesting that GUI-specific diagnosis and validation help harness improvements generalize to unseen tasks. The code is available at https://github.com/GaryYang12345/GUI-HARVEST.

cs.LG↗
arXiv2026

Analyticity Bootstrap of Feynman Integrals

We propose an analyticity bootstrap method to determine the reduction of Feynman integrals (FIs). The key observation is that, at any kinematic point, a dimensionally regularized FI possesses a Taylor expansion region (analytic region), and the finiteness of the Taylor region strongly constrains the integral reduction coefficients, which are rational functions of the kinematic variables. By writing the most general ansatz compatible with the singularity structure and then fixing the remaining parameters with only a few integration-by-parts (IBP) samples or asymptotic expansions, one can obtain the full reduction. We develop the method systematically in both single-variable and multi-variable cases, with many explicit examples. Interestingly, the differential equations (DEs) of the master integrals determine all these analyticity constraints and, conversely, the constraints facilitate the construction of the DEs. In particular, we show that the DEs in the auxiliary mass flow method can be constructed at significantly reduced cost and can even be fully determined without using IBP.

hep-ph↗
arXiv2026

Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts

The rapid advancement of large language models (LLMs) has made machine-generated text increasingly difficult to distinguish from human-written text. While recent studies explore leveraging internal representations of language models to uncover deeper detection signals, these raw features often exhibit substantial overlap between classes, limiting their discriminative power. To address this challenge, we propose Steer-to-Detect (\texttt{S2D}), a two-stage framework for detecting LLM-generated text. In the first stage, \texttt{S2D} learns a steering vector that is injected into the hidden states of a frozen observer LLM, producing representations with improved class separability. In the second stage, detection is performed via a hypothesis testing procedure based on the steered representations. We establish finite-sample, high-probability guarantees for Type I and Type II errors, providing a theoretical characterization of the procedure. Empirically, \texttt{S2D} achieves strong and consistent performance across a range of settings, including out-of-distribution scenarios and adversarial perturbations.

stat.AP↗
arXiv2026

Weak Convergence Rates for Partial-Sum Processes of Nonlinear Stochastic Approximation

Stochastic approximation provides a general framework for online estimation and optimization. Statistical inference based on the resulting estimates requires understanding their fluctuations around the target. For Polyak--Ruppert averaging, functional central limit theorems describe the normalized cumulative estimation errors through a Brownian limit. However, weak convergence alone does not determine how quickly the distribution of this process approaches that limit. We establish a quantitative functional central limit theorem for nonlinear stochastic approximation with polynomially decreasing step sizes and martingale-difference noise. Under suitable regularity and moment conditions, we derive an explicit finite-sample bound on the Prokhorov distance between each fixed scalar projection of the normalized partial-sum process and its Brownian limit. The bound reveals a step-size tradeoff: faster step-size decay reduces the contribution of the nonlinear remainder but increases that of the smoothing induced by the recursion. To assess the sharpness of the upper bound, we develop four lower-bound constructions, each designed to isolate one source of error. They show that covariance stabilization, finite-moment noise, algorithmic smoothing, and nonlinearity can each separately limit the Brownian approximation rate of the SA path.

math.ST↗