arXiv subjects

Xiaohong Chen

Publications and source records attributed to Xiaohong Chen.

At least 19 recordsLinked to original sources

arXiv2026

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation

Self-evolving agent skills enable AI agents to accumulate and refine reusable procedural knowledge from interaction experience without updating model parameters. However, since such skills are typically represented as textual documents and executed through LLM reasoning, even previously learned procedures must be repeatedly interpreted and regenerated, introducing avoidable uncertainty into every execution. In contrast, program execution is deterministic and independently verifiable. Intuitively, transforming well-specified operations into executable programs can reduce reliance on LLM reasoning and, consequently, reduce execution uncertainty. Motivated by this intuition, we introduce SKillCompiler, a self-evolving skill framework that integrates programmatic compilation into the skill evolution process. Specifically, SkillCompiler adopts an executability-aware skill compilation strategy that decomposes skills into fine-grained operations, analyzes their executability, and constructs hybrid skill skeletons that retain LLM reasoning for context-dependent operations while compiling well-specified ones into validated programs. To avoid redundant program synthesis and enable executable knowledge to accumulate across evolving skills, we further design a program asset reuse mechanism that organizes validated programs as reusable assets, retrieves existing implementations whenever applicable, and synthesizes and verifies new programs only when necessary. Extensive experiments on five benchmarks and six model families demonstrate that SkillCompiler consistently outperforms state-of-the-art methods, achieving up to 32.33\% improvement in task performance.

cs.AI↗

arXiv2026

SLIM: Stochastic Learning and Inference in Nonlinear GMM Models

We propose SLIM (Stochastic Learning and Inference in overidentified Models), a new scalable stochastic approximation algorithm for nonlinear GMMs with many nonseparable moments and parameters. SLIM is shown to be consistent and asymptotically normal without a consistent initial estimator or a globally convex GMM criterion. Random-scaling and plug-in based confidence set constructions remain valid when computational uncertainty can be non-negligible relative to sampling uncertainty. An optional second-order refinement achieves full-sample GMM efficiency. We develop $J$-tests for SLIM. The theoretical results are extended to models with clustered data of unequal cluster sizes. Monte Carlo experiments for a nonlinear demand system with 576 moments and 380 parameters demonstrate computational gains and scalability to one million observations. A LinkedIn application with 4.8 million college graduates illustrates clustered estimation and inference for first employment destinations.

econ.EM↗

arXiv2026

FedSkillGit: Federated Skill Evolution via Version Control

Large language model agents can improve their capabilities by evolving reusable natural-language skills from execution experience. In real-world deployments, this experience is distributed across data owners who cannot pool their raw data, which motivates federated skill evolution. The natural first design lets each client evolve the whole skill document and upload the result for server-side aggregation. That design is poorly matched to skill evolution for three reasons. An evolved document can still carry client-specific content distilled from private traces, repeated transmission of a nearly unchanged document spends communication on knowledge that the population already holds and the contribution of a single client is buried in a document that the server cannot meaningfully attribute. We propose FedSkillGit, a federated skill evolution framework that replaces full-document updates with commits. Each participating client submits a bounded set of patches over named rules together with its local experience, which summarizes what the round taught it, and the pair passes a two-stage privacy filter before it leaves the client. The server deduplicates the submitted patches and the local experience of all participants, updates a global experience document that only the optimizers read, and merges the surviving patches into a candidate skill. The candidate is distributed to the clients, which score it on their own validation sets and report the results, and it is published only when the reported scores improve on the current version. Experiments on three datasets and three language model families show that FedSkillGit outperforms state-of-the-art baselines by up to 22% and approaches centralized evolution without sharing raw client data.

cs.AI↗

arXiv2026

Evaluating Exact Output and Checkpoint-State Prediction in Real Programs

We present a benchmark for predicting final output and checkpoint state from source and input alone. It extends CRUXEval-style output prediction with paired shorter- and longer-trace inputs and checkpoints inside and after a loop. The benchmark contains 400 cases from 371 Python and C++ programs, evaluated under seven settings from four model families without tools or code execution. Of 11,200 planned predictions, 11,151 produced gradable responses. Reasoning-enabled settings outperform their off counterparts by 33.1 to 55.2 percentage points on completed responses. The strongest setting scores 93.0% on shorter-trace final output, 77.0% on longer-trace final output, and 65.5% and 63.5% on the two state tasks; these scores also hold when missing responses count as wrong. Across 2,397 matched Python comparisons with identical source, changing to the longer-trace input yields 528 correct-to-wrong changes and 147 reversals. Source-clustered analyses preserve this accuracy gap, while adjusted Python models give no evidence of a positive incremental association between cumulative state load and error. Changed inputs and checkpoint tasks alter several factors together, so the gaps do not isolate trace length or an internal state-tracking mechanism. The benchmark exposes errors hidden by short-output scores alone.

cs.SE↗

arXiv2026

From Verification Failures to Reusable Guidance for Coding Agents

Coding agents need to establish that a program satisfies a specification and that the specification captures the requested behavior. We study how expert diagnosis of verification failures can become reusable guidance for this work. Our approach combines executable language definitions in the K framework with a kit of procedures for constructing specifications, repairing proofs, and auditing their adequacy. A human-guided development campaign on HumanEval, a benchmark of 164 Python programming tasks, achieves a 164/164 success rate with the semantics and the kit, measured by final AI audit Pass verdicts after two targeted repairs. To examine whether auditing detects problems that successful proofs leave unresolved, we construct 12 author-reviewed pairs of clean and defective packages. Every package passes its K proofs, and completed audits identify all defects and accept all clean packages. We then use KleverBench to test specification and proof construction for 31 programs with changed operator meanings. Comparisons with complete acceptance rules and equally long generic advice yield mixed results across two model and budget settings, motivating further work on selecting useful guidance within resource limits. Human-reviewed Optimism proofs establish expected pause reverts for six operations within declared input bounds under London semantics with unbounded gas. We report progress, difficulties, and lessons toward agents that deliver programs with checkable correctness arguments.

cs.SE↗

arXiv2026

Repair or Resample? Rethinking Failure Debugging in LLM Multi-Agent Systems

As large language model (LLM)-based multi-agent systems (MASs) are increasingly applied to long-horizon complex tasks, their reliability has emerged as the core bottleneck hindering their real-world deployment. Existing MAS debugging and repair methods typically rely on rerunning and resampling the entire execution trajectory. However, a fundamental question remains to be answered: do these methods causally repair MAS failures or merely stochastically repair by leveraging the randomness of LLM sampling? To evaluate the effectiveness of MAS repair methods, we introduce SymTrace, a controlled evaluation framework that records the MAS execution trajectory and establishes intervention anchors. During replay, it effectively reconstructs the execution before the anchor using recorded logs and only regenerates the downstream trajectory, thereby enabling the reliable reproduction of MAS failures. We further construct the dataset SymFail, comprising 536 human-annotated failure trajectories with graph-linked locations, categories, and trace evidence. Based on these foundations, we conduct a large-scale empirical study across three mainstream MAS frameworks. Our findings reveal that existing unguided rerun methods are highly unreliable, exhibiting low failure reproduction and repair rates (only 67.97% and 6.90%, respectively). Building upon these findings, we further explore the effectiveness of a symptom-driven intervention method, which successfully repairs 20.15% of the failed cases (a 191.89% improvement to state-of-the-art repair methods). This study aims to provide actionable insights for MAS debugging and repair research, paving the way for the robust deployment of multi-agent systems.

cs.AI↗

arXiv2026

Offline Deep Q* Estimation with Diffusion Models

In offline RL, estimating the optimal action-value function $Q^*$ can be formulated as solving the optimal Bellman equation based solely on offline observations. A fundamental challenge is that the reward function and transition kernel are unknown, so the optimal Bellman operator is not directly observable from data. To address this issue, we propose a novel framework that decouples operator estimation from value function learning. In this approach, we first formulate conditional diffusion models to estimate the reward law and transition kernel, which induces a data-driven approximation of the optimal Bellman operator. We then plug these estimators into the Bellman equation and obtain a deep estimator of $Q^*$ by minimizing the empirical Bellman residual over a neural network function class. Theoretically, we first establish sharp nonasymptotic convergence rates for learning the optimal Bellman operator through an end-to-end analysis of conditional diffusion estimation in total variation distance. We then establish the oracle value-stage rate $\widetilde{\mathcal O}\bigl(n^{-\frac{2β}{d_x+d_a+2β}}\bigr)$ for the excess Bellman residual risk. Finally, under a concentrability condition, we translate this residual bound into an $L^2$ convergence rate of $\widetilde{\mathcal O}\bigl(n^{-\fracβ{d_x+d_a+2β}}\bigr)$ for the resulting deep estimator of $Q^*$, where $d_x$ and $d_a$ denote the dimensions of the state and action spaces, respectively, and $β$ denotes the Hölder smoothness index of $Q^*$. Importantly, our theoretical analysis does not rely on completeness assumptions commonly used in deep RL theory. Extensive numerical experiments demonstrate the effectiveness of the proposed method and its strong empirical performance.

stat.ML↗

arXiv2026

Completeness and incompleteness of basic matching logic

Basic matching logic is matching logic without definedness. Symbols are interpreted as set-valued operations, element variables denote singletons and are bound by $\exists$, and no connective uniformly internalizes totality. For basic matching logic without fixpoints over an arbitrary one-sorted finitary signature, we prove \emph{global completeness} ($Γ\vDashφ$ iff $Γ\vdashφ$, for arbitrary, possibly infinite $Γ$) and, as a corollary, conservativity of the definedness extension. The proof localizes $Γ$ to a theory $Δ_Γ$ and reduces semantic consequence and derivability to the same local relation: $Γ\vDashφ$ iff $Δ_Γ\vDash_\text{loc}φ$ iff $Γ\vdashφ$. A double-cover construction establishes the semantic equivalence. Least fixpoints destroy effective axiomatizability. Over a signature with one unary and two binary symbols and no constants, validity is not recursively enumerable; hence no sound calculus with a recursively enumerable proof relation is even weakly complete, already for the empty theory and without definedness. The positive result is also sharp in the number of sorts. Global completeness fails with three sorts for a satisfiable $Γ$. Thus the completeness conjecture holds for one sort and fails in general. The negative results arise from sort flow, fixpoint effectivity, and, for hybrid logic, an obstruction to every well-founded calculus whose leaves are hypotheses or valid patterns and whose rules respect localization. This yields a matching-logic-independent dichotomy: the language with state variables bound by $\exists$ and $\forall$ over modalities of arbitrary arity is globally complete without nominals, while no calculus in that well-founded class is globally complete once nominals are added.

cs.LO↗

arXiv2026

Fast Online Inference on Semiparametric Models

This paper develops a framework for fast online inference on semiparametric models with large sample sizes and possibly many covariates. The computational algorithm itself is the object of statistical study: after a globally consistent warm start in the first phase, the path of averaged online iterates generated in the second phase automatically delivers estimators with optimal convergence rates and valid confidence sets. Both phases require only a single pass over the data stream and are well suited to streaming data or to settings with storage/privacy constraints. For semiparametric monotone index models, the averaged trajectory of the second phase lead to estimators that are automatically orthogonalized and satisfy the laws of the iterated logarithms, and policy functionals are updated along the same trajectory at negligible additional cost. The averaged trajectories satisfy functional central limit theorems, which yield fast online inference via random scaling and bypass the explicit variance estimation that complicates inference for semiparametric models. Applied to a fixed large sample, our online algorithm achieves substantial computational gains over corresponding offline procedures without sacrificing statistical performance. Monte Carlo experiments show adequate behavior. Our methods are applied to 19 million traffic-stop online records from the North Carolina State Patrol (Pierson et al. 2020) and to the international trade data of Helpman et al. (2008) with over 300 regressors. We also compare our estimator to its parametric benchmark in both empirical illustrations

econ.EM↗

arXiv2026

Partial Identification from LLM Prompts

Large language models are increasingly used as binary classifiers when the true label is latent. We study partial identification of the prevalence $θ= P(X^* = 1)$ from panels of LLM reports whose errors may be arbitrarily dependent given the truth. The design of replication determines the observable, and hence the identifying content: repeated prompts to one model yield a count, several named models a response vector, and both a response matrix. Cast as a two-component finite mixture, the problem makes the identification failure transparent: absent restrictions that separate the latent components, the prevalence $θ$ is completely unidentified, and weak stochastic-ordering restrictions (first-order dominance, monotone likelihood ratio, mean ordering) leave the identified set at $[0,1]$. Identifying power comes instead from externally calibrated scores and events, which discipline the mixture in the spirit of the misclassification and corrupted-data literature. We characterize the resulting bounds, establishing validity and sharpness, and give an exact account of the identifying information in the full score distribution beyond its mean. When named models are asked repeated versions of the same question, what identifies $θ$ is not the number of positive answers but which models agree across prompts -- a feature a vote count discards. An extension derives implied bounds on regression coefficients when $X^*$ is a regressor of interest that is not directly observed.

econ.EM↗

arXiv2026

Exploring and Complementing End Users' Requirements in IoT enabled System

End users create IoT automation rules via trigger action programming, but their expressions are often fragmented, capturing device operations rather than high level intents. This gap leads to missing conditions, logical conflicts, and overlooked safety constraints, risking hazardous behaviors. To address this, we propose an intent driven requirements completion approach that reframes rule completion as a dual process: reconstructing intent from fragmented rules, then regenerating rules from that intent, with safety embedded throughout. We introduce a Bidirectional Requirements Traceability Tree, a three layer model linking rules, intents, and quality concerns, and design a multiagent framework that combines LLM reasoning with structured traceability. This enables completions that are both functionally complete and inherently safe, while remaining traceable and explainable. Evaluation shows our method significantly outperforms the baselines, improving the rule completion rate by 43% and reducing logical conflicts by over 21%. By grounding completion in intent understanding, we shift the paradigm from user to system responsibility, and from functional correctness to holistic trustworthiness.

cs.SE↗

arXiv2026

From Chat to Interview: Agentic Requirements Elicitation with an Experience Ontology

Requirements elicitation interviews are crucial and time-consuming in requirements engineering, but heavily rely on the experience of requirements analysts. Although recent advancements in large language models (LLMs) have created new opportunities to automate this process, existing approaches rely solely on LLMs for free-form chat without taking into account the interview and development experience. That leads to the omission of implicit requirements and redundant questions. Practically, experienced analysts implicitly follow a structured cognitive framework when conducting requirements elicitation. Inspired by this observation, this paper proposes an interview agent named OntoAgent for the elicitation of requirements guided by an experience ontology. OntoAgent automatically analyzes domain-specific requirements descriptions to construct an experience ontology, which organizes requirements concerns into an ontology to support systematic and explainable interviews. During the interview, OntoAgent first performs four operations (i.e., ParseUser, ScoreOnto, ReRankOnto, GatePrune) guided by the ontology to identify the relevant requirement concerns. The selected concern is then combined with the current dialogue context to generate the elicitation question. To validate OntoAgent, we conduct comprehensive quantitative experiments using the widely adopted website application domain. The results show that OntoAgent significantly outperforms existing baselines in both elicitation effectiveness and questioning efficiency, achieving a 33% improvement in IRE and a 21% improvement in TKQR. Ablation studies further validate the contribution of each key design component. In addition, a qualitative user study demonstrates its practical advantages in real-world scenarios. We believe that OntoAgent can also be extended to requirements interview tasks in other domains.

cs.SE↗

arXiv2026

ClarifySTL: An Interactive LLM Agent Framework for STL Transformation through Requirements Clarification

Signal Temporal Logic (STL) is a formal language for specifying real-time behaviors of cyber-physical systems (CPS). Automatically transforming natural language requirements into STL specifications has received growing attention. Recent efforts leveraging large language models (LLMs) have demonstrated impressive performance, but some natural language requirements in practice contain vague or ambiguous information, which remains challenging for LLMs to handle. To address these challenges, we propose ClarifySTL, an interactive LLM-agent framework that enhances STL transformation through requirements clarification. ClarifySTL first detects vague expressions that indicate underspecified information in a requirement. If any vagueness is detected, it generates targeted clarification queries to guide users in supplementing the requirement until all necessary details are provided. Subsequently, if ClarifySTL detects ambiguities, it formulates focused ambiguity clarification queries and updates the requirements based on user feedback until all ambiguities are resolved. Finally, the requirements with vagueness and ambiguity clarified are transformed into STL specifications using LLMs. This interactive framework ensures that the resulting STL formulas faithfully capture user intent while reducing the burden on the user. We evaluate ClarifySTL on the representative benchmarks DeepSTL and STL-DivEn, as well as our newly introduced AmbiEval benchmark, which is specifically designed to assess the performance of the agents in handling vagueness and ambiguity, including both detection and query generation. The experimental results show that ClarifySTL is effective.

cs.SE↗

arXiv2026

Thin Sets Are Not Equally Thin: Minimax Learning of Submanifold Integrals

Many economic parameters are identified by ``thin sets'' (submanifolds with Lebesgue measure zero) and hence difficult to recover from data in an ambient space. This paper provides a unified theory for estimation and inference of such ``thin-set'' identified functionals. We show that thin sets are \emph{not} equally thin: their intrinsic dimensionality $m$ matters in a precise manner. For a nonparametric regression $h_0$ with Hölder smoothness $s$ and $d$-dimensional covariates in the ambient space, we show that $n^{-\frac{s}{2s+d-m}}$ is the minimax optimal rate of estimating linear and nonlinear (e.g., quadratic, upper contour) integrals of $h_0$ on an $m$-dimensional submanifold ($0\leq m < d$), which is the fastest possible attainable rate among all estimators. The minimax lower bound rate result is generalized to estimating submanifold integrals when $h_0$ is a nonparametric density and a nonparametric instrumental variable function. The asymptotic normality of t statistics is established via sieve Riesz representation, and the corresponding inference is computed using Sobol points.

econ.EM↗

arXiv2026

On Local Overidentification and Efficiency Gains in Modern Causal Inference and Data Combination

This paper studies nonparametric local (over-)identification and the semiparametric efficiency in modern causal frameworks. We develop a unified approach that begins by translating structural models with latent variables into their induced statistical models of observables and then analyzes local overidentification through conditional moment restrictions. We apply this approach to three popular classes of causal models: (1) the general treatment model under unconfoundedness; (2) the negative control model, and (3) the long-term causal inference model under unobserved confounding. The first model yields a locally just-identified statistical model, implying that all regular asymptotically linear estimators of the treatment effect have the same asymptotic variance, which equals the (trivial) semiparametric efficient variance bound. In contrast, the latter two models involve nonparametric endogeneity and are naturally locally overidentified; consequently, some doubly robust orthogonal moment estimators of the average treatment effect are inefficient. Whereas existing work typically imposes strong conditions to restore local just-identification to justify the efficiency of their doubly robust orthogonal moment estimators, we characterize the semiparametric efficient variance bounds, along with efficient estimators, for the (locally) overidentified models (2) and (3). A small real data application, along with a simulation study, illustrates the semiparametric efficiency gains in model (3).

econ.EM↗

arXiv2026

Sampling from Constrained Gibbs Measures: with Applications to High-Dimensional Bayesian Inference

This paper considers a non-standard problem of generating samples from a low-temperature Gibbs distribution with \emph{constrained} support, when some of the coordinates of the mode lie on the boundary. These coordinates are referred to as the non-regular part of the model. We show that in a ``pre-asymptotic'' regime in which the limiting Laplace approximation is not yet valid, the low-temperature Gibbs distribution concentrates on a neighborhood of its mode. Within this region, the distribution is a bounded perturbation of a product measure: a strongly log-concave distribution in the regular part and a one-dimensional exponential-type distribution in each coordinate of the non-regular part. Leveraging this structure, we provide a non-asymptotic sampling guarantee by analyzing the spectral gap of Langevin dynamics. Key examples of low-temperature Gibbs distributions include Bayesian posteriors, and we demonstrate our results on three canonical examples: a high-dimensional logistic regression model, a Poisson linear model, and a Gaussian mixture model.

math.ST↗

arXiv2026

ReqElicitGym: An Evaluation Environment for Interview Competence in Conversational Requirements Elicitation

With the rapid improvement of LLMs' coding capabilities, the bottleneck of LLM-based automated software development is shifting from generating correct code to eliciting users' requirements. Despite growing interest, the interview competence of LLMs in conversational requirements elicitation remains fully underexplored. Existing evaluations often depend on a few scenarios, real user interaction, and subjective human scoring, which hinders systematic and quantitative comparison. To address these challenges, we propose ReqElicitGym, an interactive and automatic evaluation environment for assessing interview competence in conversational requirements elicitation. Specifically, ReqElicitGym introduces a new evaluation dataset and designs both an interactive oracle user and a task evaluator. The dataset contains 101 website requirements elicitation scenarios spanning 10 application types. Both the oracle user and the task evaluator achieve high agreement with real users and expert judgment. Using our ReqElicitGym, any automated conversational requirements elicitation approach (e.g., LLM-based agents) can be evaluated in a reproducible and quantitative manner through interaction with the environment. Based on our ReqElicitGym, we conduct a systematic empirical study on seven representative LLMs, and the results show that current LLMs still exhibit limited interview competence in uncovering implicit requirements. Particularly, they elicit less than half of the users' implicit requirements, and their effective elicitation questions often emerge in later turns of the dialogue. Besides, we found LLMs can elicit interaction and content implicit requirements, but consistently struggle with style-related requirements. We believe ReqElicitGym will facilitate the evaluation and development of automated conversational requirements elicitation.

cs.SE↗

arXiv2026

REprompt: Prompt Generation for Intelligent Software Development Guided by Requirements Engineering

The rapid development of large language models is transforming software development. Beyond serving as code auto-completion tools in integrated development environments, large language models increasingly function as foundation models within coding agents in vibe-coding scenarios. In such settings, prompts play a central role in agent-based intelligent software development, as they not only guide the behavior of large language models but also serve as carriers of user requirements. Under the dominant conversational paradigm, prompts are typically divided into system prompts and user prompts. System prompts provide high-level instructions to steer model behavior and establish conversational context, while user prompts represent inputs and requirements provided by human users. Despite their importance, designing effective prompts remains challenging, as it requires expertise in both prompt engineering and software engineering, particularly requirements engineering. To reduce the burden of manual prompt construction, numerous automated prompt engineering methods have been proposed. However, most existing approaches neglect the methodological principles of requirements engineering, limiting their ability to generate artifacts that conform to formal requirement specifications in realistic software development scenarios. To address this gap, we propose REprompt, a multi-agent prompt optimization framework guided by requirements engineering. Experiment results demonstrate that REprompt effectively optimizes both system and user prompts by grounding prompt generation in requirements engineering principles.

cs.SE↗