arXiv subjects

Nan Jiang

Publications and source records attributed to Nan Jiang.

At least 19 recordsLinked to original sources

arXiv2026

Q-MMR: Off-Policy Evaluation via Recursive Reweighting and Moment Matching

We present a novel theoretical framework, Q-MMR, for off-policy evaluation in finite-horizon MDPs. Q-MMR learns a set of scalar weights, one for each data point, such that the reweighted rewards approximate the expected return under the target policy. The weights are learned inductively in a top-down manner via a moment matching objective against a value-function discriminator class. Notably, and perhaps surprisingly, a data-dependent finite-sample guarantee for general function approximation can be established under only the realizability of $Q^π$, with a dimension-free bound -- that is, the error does not depend on the statistical complexity of the function class. We also establish connections to several existing methods, such as importance sampling and linear FQE. Further theoretical analyses shed new light on the nature of coverage, a concept of fundamental importance to offline RL.

cs.LG↗

arXiv2026

ViT-H: Generative Video-To-Humanoid Control with Motion Tracking Feedback

Learning humanoid control from monocular human videos offers a scalable way to acquire diverse whole-body behaviors. Yet, existing reconstruction-retargeting-tracking pipelines are limited in system-level scalability by separately optimized intermediate stages and cannot directly optimize reference motion executability using downstream tracking feedback. We present ViT-H, a video-to-humanoid control framework following a generation-and-tracking paradigm. ViT-H directly generates humanoid reference motions from videos with a diffusion model, bypassing human motion reconstruction and retargeting, and uses motion-tracking outcomes to post-train the generator with direct preference optimization. We further introduce an autoregressive variant that enables real-time humanoid video teleoperation with sub-400 ms latency. Extensive experiments on the Unitree G1 demonstrate that ViT-H outperforms state-of-the-art reconstruction-retargeting-tracking pipelines in both motion quality and executability. We further show that ViT-H scales favorably with model size and with both pre-training and post-training data scale, while motion-tracking feedback consistently improves executability across different motion trackers.

cs.RO↗

arXiv2026

WAFFLE: Finetuning Multi-Modal Models for Automated Front-End Development

Web development involves turning UI designs into functional webpages, which can be difficult for both beginners and experienced developers due to the complexity of HTML's hierarchical structures and styles. While Large Language Models (LLMs) have shown promise in generating source code, two major challenges persist in UI-to-HTML code generation: (1) effectively representing HTML's hierarchical structure for LLMs, and (2) bridging the gap between the visual nature of UI designs and the text-based format of HTML code. To tackle these challenges, we introduce Waffle, a new fine-tuning strategy that uses a structure-aware attention mechanism to improve LLMs' understanding of HTML's structure and a contrastive fine-tuning approach to align LLMs' understanding of UI images and HTML code. Models fine-tuned with Waffle show up to 9.00 pp (percentage point) higher HTML match, 0.0982 higher CW-SSIM, 32.99 higher CLIP, and 27.12 pp higher LLEM on our new benchmark WebSight-Test and an existing benchmark Design2Code, outperforming current fine-tuning methods.

cs.SE↗

arXiv2026

Can Language Models Replace Programmers for Coding? REPOCOD Says 'Not Yet'

Recently, a number of repository-level code generation benchmarks-such as CoderEval, DevEval, RepoEval, RepoBench, and LongCodeArena-have emerged to evaluate the capabilities of large language models (LLMs) beyond standalone benchmarks like HumanEval and MBPP. Thus, a natural question is, would LLMs have similar performance in real world coding tasks as their performance in these benchmarks? Unfortunately, one cannot answer this question, since these benchmarks consist of short completions, synthetic examples, or focus on limited scale repositories, failing to represent real-world coding tasks. To address these challenges, we create REPOCOD, a Python code-generation benchmark containing complex tasks with realistic dependencies in real-world large projects and appropriate metrics for evaluating source code. It includes 980 whole-function generation tasks from 11 popular projects, 50.8% of which require repository-level context. REPOCOD includes 314 developer-written test cases per instance for better evaluation. We evaluate ten LLMs on REPOCOD and find that none achieves more than 30% pass@1 on REPOCOD, indicating the necessity of building stronger LLMs that can help developers in real-world software development. In addition, we found that retrieval-augmented generation achieves better results than using target function dependencies as context.

cs.SE↗

arXiv2026

TENET: One Step Toward Test-Driven Development for Repository-Level Code Generation

Test-Driven Development (TDD) is a widely adopted practice that requires developers to create and execute tests alongside implementation. With recent advances in Large Language Models (LLMs), developers can shift from manually writing the code to defining tests as executable specifications and delegating code synthesis to AI agents. However, enabling repository-level TDD under developer-written tests is challenging, requiring: (1) specification enhancement: identifying a concise yet representative test subset from large suites with rich task semantics; (2) retrieval augmentation: using tests to guide reasoning and context retrieval; and (3) test-driven refinement: interpreting test feedback for iterative improvement. We propose TENET, an agentic framework for repository-level code generation under the TDD paradigm. TENET includes: (1) a test harness mechanism that selects a concise test suite to maximize diversity of the target usage scenarios; (2) a tailored agent toolset for efficient retrieval and debugging; and (3) a reflection-based refinement workflow that iteratively analyzes failures and updates implementations. TENET consistently outperforms the strongest baselines across backbones, achieving 69.08% and 81.77% Pass@1 on RepoCod and RepoEval with Claude Sonnet 4, improving by 9.49 and 2.17 percentage points, respectively. Additionally, we present the first systematic study of how test suite characteristics influence LLM agent performance in TDD settings.

cs.SE↗

arXiv2026

Aligning Language Models Towards Counterfactual Invariance

Small variations in prompt characteristics can meaningfully alter the semantics, tone, and contextual assumptions of large language model outputs. Such characteristics can involve group variables such as demographics or other sensitive latent attributes. Sensitivity to such input variables is undesirable when demographic information is causally irrelevant to the task, but might still be relevant for generation of semantically coherent responses. This makes the task of enforcing counterfactual invariance while preserving model utility difficult. To study this problem, we first show that standard model alignment recipes (such as reinforcement learning) increase differences in response quality across demographics, while improving overall response quality, ultimately amplifying what is known as counterfactual bias. To address this, we develop a policy gradient technique involving a group-level reward function that operates on matched pairs of counterfactual prompts. For completeness, we build a dataset of controlled counterfactual prompt pairs and design an evaluation that separately measures cases where demographic information is causally relevant to the answer. Finally, we validate our findings on several fairness benchmarks. We demonstrate that our methods reduce counterfactual differences while preserving overall response quality and that the models continue to adapt appropriately when such sensitive information is relevant.

cs.CL↗

arXiv2026

ARGUS: An Agenda-Driven Framework for Reliable User Simulation

High-fidelity user simulators are essential for evaluating and training agents in interactive environments. However, current LLM-based simulators struggle with reliability issues, such as goal omission and policy non-compliance. We propose ARGUS (Agenda-dRiven Goal-oriented User Simulator), a neuro-symbolic framework that maintains the user's goal state using a dynamic agenda stack manipulated via symbolic operations. By rigorously decoupling state tracking from natural language generation, ARGUS facilitates strict goal progression and constraint-aligned interactions. For cost-efficient deployment, we distill ARGUS into compact models using a unified pipeline: evaluator-guided SFT followed by online RL with dense step-level rewards. To rigorously evaluate behavioral fidelity, we introduce a rubric-based metric Composite Behavioral Score (CBS). Evaluations on $τ^3$-Bench and MultiWOZ demonstrate the significant superiority of the ARGUS architecture and the efficacy of our training pipeline. Code and data will be released soon.

cs.AI↗

arXiv2026

Analyzing and Mitigating Cost-Inefficient Behaviors in Coding Agents

Although effective, coding agents often incur substantial monetary costs. Their recurring cost-inefficient behaviors remain underexplored. We conduct the first study of behavioral cost inefficiencies in coding agents, analyzing 1,200 trajectories from Claude Code and Mini-SWE-Agent across four configurations on SWE-bench Verified. We identify three cost-inefficient behaviors: subsumed retrieval, similar script generation, and test re-execution. We then evaluate three mitigation strategies: structure-aware retrieval, agent-synthesized skills, and developer-designed skills, over 10k trajectories on held-out SWE-bench Verified and Pro tasks. Our main findings are: (1) The three behaviors affect 79.00%--98.00% of coding tasks and account for up to 22.75% of task cost. (2) Structure-aware retrieval can introduce retrieval overhead and alter agent delegation, causing inconsistent improvements in retrieval efficiency and cost increases of up to 28.14%. (3) Agent-synthesized skills tend to produce low-level, trace-specific guidance, limiting their effectiveness and generality. (4) In contrast, developer-designed skills provide high-level, trace-agnostic guidance, reducing cost by up to 41.73%, roughly twice the maximum gain from agent-synthesized skills.

cs.AI↗

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

PreSTO: Predictive Subtree Prefetching for Fast LLM Power Sampling

Power sampling improves large language model (LLM) reasoning at test time by sampling from a sharpened version of the base model's distribution. Its sampler, PowerMH, is slow because its Metropolis--Hastings (MH) updates are sequential: each accept/reject decision sets the state from which the next proposal is generated. Prior work reduces this cost by changing the sampler or the model (e.g., sequential Monte Carlo, approximate lookahead, distillation). We propose Predictive SubTree Prefetching (PreSTO), which accelerates PowerMH by prefetching future proposals and provably leaves its MH transition kernel unchanged. PreSTO views future MH outcomes as a tree of accept/reject decisions: in each model call, it generates the proposals of a \textit{prefetched subtree} of this tree in one batch, then follows the realized path with the original MH acceptance rule. We also observe that MH reaches only a sparse, nearly unary part of each prefetched subtree, we further design a \textit{predictive traversal rule} that chooses the subtree to maximize the predicted number of completed transitions. Across seven datasets and five base LLMs, PreSTO achieves median speedups of $1.71\times$, $1.91\times$, and $2.33\times$ on PowerMH and its follow-ups EntropyCut and MultiTryMH, respectively; on the first two, it makes a median of $3.9\times$ fewer model calls, each only $2.1\times$ longer. On LCB V6 with Qwen3.5-9B, the predictive traversal rule is a further $1.25\times$--$1.37\times$ faster than the default breadth-first traversal.

cs.CL↗

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

DecomVoxel: Harnessing 3D-Native Priors with Guided In-situ Denoising Optimization for Decompositional Scene Reconstruction

Decompositional scene reconstruction aims to reconstruct high-quality objects and background, yet existing methods still struggle with the level of quality under heavy occlusions. While generative priors offer a potential solution, 2D image-based priors often suffer from multi-view inconsistency due to a lack of 3D awareness. Conversely, 3D-native priors provide stronger structural inductive biases but frequently lead to spatial drift and misalignment within complex scenes. To address these issues, we propose DecomVoxel, formulating object completion as a guided in-situ denoising optimization that bridges 3D-native priors with neural scene reconstruction. Our framework introduces a reformulated epsilon-based distillation loss to ensure stable latent refinement, alongside adaptive spatial guidance that utilizes occupied and vacant anchors with temporal annealing to suppress generative hallucinations and mitigate spatial drift. Experiments on Replica and ScanNet++ show that DecomVoxel significantly outperforms state-of-the-art methods while faithfully preserving the original spatial layout, structural fidelity, and style-consistent texture. Our method pushes the boundary of decompositional reconstruction by delivering high-quality textured meshes with clean topology, geometry, and appearance, providing a robust solution for the decompositional reconstruction of complex real-world scenes. Code is available at https://github.com/DecomVoxel/DecomVoxel.

cs.CV↗

arXiv2026

Explore Broadly, Reason Sharply: Push Small Models toward the Frontier via Sampling

Power-sharpened sampling is an inference-time alternative to reinforcement-learning (RL) post-training for enhancing reasoning in large language models (LLMs). High-probability sequences are amplified under the base model without parameter updates or external rewards, avoiding the costly optimization and jagged generalization of RL. However, this approach faces a fundamental exploration--exploitation trade-off, as % strong sharpening restricts exploration, trapping samplers in plausible but incorrect reasoning trajectories, whereas weak sharpening leaves the answer distribution diffuse. To resolve this trade-off, we introduce \textbf{Parallel Power Tempering (PPT)}, instantiating power-sharpened LLM sampling via parallel tempering. Running multiple \emph{interacting} replicas in parallel at different sharpening levels allows lower-power replicas to explore diverse reasoning trajectories and higher-power chains to further exploit higher-likelihood responses favored by the sharpened target. Specifically, we tailor \method{} to inference-time sampling by mitigating a truncation bias, identified in prior power samplers, and investigate effective swap strategies under finite memory and compute budgets. Extensive experimentation shows that \method{} substantially improves single-chain power-sharpened sampling and outperforms RL-post-trained models, producing higher-quality reasoning traces and even achieving performance comparable to frontier models.

cs.LG↗

arXiv2026

Contextual Distributionally Robust Chance-Constrained Programs: Exact Reformulations and Valid Inequalities

We study a contextual distributionally robust chance-constrained programming (C-DRCCP) model that controls violation probabilities conditional on a neighborhood of a target context. The ambiguity set is defined over the joint distribution of contexts and outcomes using the Wasserstein distance. For affine safety systems with right-hand-side uncertainty, we derive an exact mixed-integer programming (MIP) reformulation. The reformulation reveals that worst-case conditional risk is determined by two coupled uses of the transportation budget: moving empirical mass into the conditioning neighborhood and moving conditioned mass into the failure region. Equivalently, robust feasibility holds when the minimum transportation cost needed to reach the conditional risk boundary is at least the available budget. Calibrating the ambiguity set gives a finite-sample guarantee for the true conditional violation probability. This characterization leads to strong valid inequalities that exploit the contextual transport. Each contextual mass allocation determines both the conditional scenario weights and the budget available to move outcomes into failure, whereas conventional quantile cuts utilize only the induced weights. Accounting for the remaining budget generates strengthened quantile cuts that strictly dominate standard quantile bounds and yield tighter MIP coefficients. We also derive rank inequalities from an exchange property of contextual transportation costs, linking feasible failure patterns across contextual allocations. Extensive numerical experiments demonstrate that these inequalities close much of the linear-relaxation gap, speed up the solution process, and enable large-sample instances that are intractable for the baseline formulation.

math.OC↗

arXiv2026

MaD-RL: Matching Distributions for Calibrating LLMs with Reinforcement Learning

Reinforcement learning (RL) is widely used in language-model post-training to maximize rewards assigned to individual model outputs, such as scores from binary verifiers or reward models trained on human feedback. However, applications such as synthetic-data generation, fairness-related constraint satisfaction, and policy exploration require controlling the distribution of outputs across model generations rather than only maximizing expected reward. We propose a general RL-based framework for \textit{Distribution Matching} allowing matching the distribution of a latent categorical attribute of model outputs to a specified target distribution. Empirically, we demonstrate that dominant post-training recipes such as Group Relative Policy Optimization (GRPO) reduce output diversity by concentrating policy probability towards a single mode. Entropy regularization and sampling temperature can improve the spread of the distribution but have constrained effectiveness, limited to apply only in token space and toward uniform distributions. We show that prior work in this area is a specific case of Distribution Matching involving the $L_2$ divergence. We then propose reward functions for other divergences such as KL and Jensen-Shannon and motivate them with theoretical justification. Finally, we demonstrate the effectiveness of our approach on a set of experiments involving mathematical reasoning and programming.

cs.LG↗

arXiv2026

SN 2025fhm: A central-engine powered Ic-BL supernova associated with X-ray transient EP250304a

We present X-ray, optical, and radio follow-up observations of EP250304a, an extragalactic fast X-ray transient (EFXT) discovered by the Einstein Probe. Its X-ray light curve exhibits two broad pulses with comparable peak fluxes within the first $\sim$1~ks, a feature rarely seen among low-luminosity gamma-ray bursts or EFXTs. Optical follow-up observations were carried out using the Korea Microlensing Telescope Network, the Thai Robotic Telescope, the Las Cumbres Observatory 1~m global network, the Gemini Multi-Object Spectrograph on Gemini south telescope, and the Global Supernova Network. The fast-cooling phase (within 3 days) of optical data can be well fitted by a shocked cocoon model. However, during the supernova phase (SN 2025fhm, from 3 to 88 days), the late-time light curve cannot be explained solely by radioactive $^{56}$Ni decay, as demonstrated by a grid of simulations using the one-dimensional Lagrangian radiation hydrodynamics code SNEC, which reveals a significant energy excess at late epochs. To account for this excess, a central engine like a rapidly spinning, highly magnetized neutron star is needed to provide additional energy injection. This model yields a best-fit spin period of $\sim$12.60~ms and magnetic field strength of $\sim 3.52\times10^{15} \rm G$, and it successfully explains both the late-time bolometric light curve and the early X-ray pulse structures. Our results indicate that EP250304a/SN 2025fhm is likely powered by a central magnetar rather than by radioactive decay alone, offering new insights into the energy budget and physical origin of EFXTs and their associated supernovae.

astro-ph.HE↗

arXiv2026

SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions

Compressing large-scale neural networks is essential for deploying models on resource-constrained devices. Most existing methods adopt weight pruning or low-bit quantization individually, often resulting in suboptimal compression rates to preserve acceptable performance drops. We introduce a unified framework for simultaneous pruning and low-bit quantization via Bayesian variational learning (\method), which achieves higher compression rates than prior baselines while maintaining comparable performance. The key idea is to employ a spike-and-slab prior to induce sparsity and model quantized weights using Gaussian Mixture Models (GMMs) to enable low-bit precision. Due to the intractability of the objective involving spike-and-slab priors with GMMs, we derive an efficient approximation that facilitates effective compression with minimal accuracy loss. In theory, we provide a consistent result for our proposed variational approach to a sparse and quantized deep neural network. Extensive experiments on compressing ResNet, BERT-base, Llama3.2, and Qwen2.5 models show that our method achieves higher compression rates than a line of existing methods with comparable performance drops. Project page: https://comeusr.github.io/SQS_Webpage.

cs.LG↗

arXiv2026

Data Assimilation with Sparse Observations

Data assimilation by nudging (also called CDA) yields exponentially decaying errors and an infinite predictability horizon if the method parameter is large enough and the observations are frequent enough in time and dense enough in space. We consider the complementary case of moderate parameters and sparse and infrequent observations. We prove that assimilation with any data and any(positive) parameter strictly decreases errors and strictly increases the (now finite) predictability horizon.

math.NA↗