Searcharxiv⌕ Search

arXiv subjects

Zixiao Wang

Publications and source records attributed to Zixiao Wang.

At least 19 recordsLinked to original sources
arXiv2026

Forecasting How RLHF Responds to Reward-Model Edits Before Training

Responses with headers, lists and greater length receive high scores from reward models (RMs), and RLHF policies learn to produce these features even when true quality does not improve. Practitioners therefore edit their RMs, but they learn whether an edit works only after an RL run. We show that the response of a policy to an RM edit can be forecast before training, from samples of the current policy and the scores of the old and new RM. If we read the RM as a set of prices, the first-order theory of KL-regularized RL implies that each feature moves in proportion to its within-prompt covariance with the change in reward. We implant controlled price changes into retrained RMs under a 2^(4-1) fractional factorial design with a self-distilled placebo, run 54 RL runs from six parent policies, and fix every forecast before the runs. With the covariance predictor, we predict the direction of all 20 changes in supply whose forecast has the same sign at every parent and explain 82% of the standardized change. With counterfactual prices, the classical alternative, we explain 50%. The forecast also holds under GRPO and for real RM upgrades, where we explain 66% out of sample with a coefficient frozen from the factorial runs. For rent, the gap between the proxy reward and a gold reward from stronger judges, we rank RM edits correctly with the same covariance over the first 30 steps (Spearman 0.94). This measures the horizon of the forecast: by step 200, the ranking of the factorial edits reverses under both RLOO and GRPO, while the real upgrades keep their order. When we vary the KL coefficient beta, the initial response does not depend on beta, and the response saturates below the KL equilibrium. In one experiment, the long-run response to a price rise is consistent with an additional implicit regularizer beta_0 of about 0.003 to 0.007.

cs.LG↗
arXiv2026

A plug-in estimator of the non-iterative conditional expectation g-formula using long short-term memory models

The non-iterative conditional expectation (NICE) plug-in estimator of the g-formula can be used to estimate the causal effects of sustained treatment strategies, even in the presence of treatment-confounder feedback, under the assumptions of exchangeability, positivity, and consistency. In practical applications, parametric models are commonly used to estimate the joint distribution of confounders, treatments, and outcomes over time. As an alternative to parametric models, long short-term memory (LSTM) models, designed to capture both short-term patterns and long-range dependencies in sequential data, may better accommodate complex temporal dependencies among time-varying confounders, treatments, and outcomes. However, the validity of statistical inferences of a plug-in estimator based on LSTM models has not yet been investigated. Here, we provide proof of principle that, under a complex data generation mechanism that resembles data from a cohort of persons with HIV and sufficiently large sample sizes, an LSTM-based NICE g-formula plug-in estimator results in less bias than parametric estimators while achieving near-nominal coverage for 95\% confidence intervals. This work is intended to spur the development of a theoretical framework that encompasses deep learning plug-in estimators for causal inference based on high-dimensional longitudinal data.

stat.ME↗
arXiv2026

BioXArena: Benchmarking LLM Agents on Multi-Modal Biomedical Machine Learning Tasks

Large language model (LLM) agents increasingly automate machine-learning model building, yet biomedical benchmarks emphasize question answering, reasoning, and tool use or cover narrow slices of biomedical ML coding. We introduce BioXArena, a biomedical machine learning (BioML) coding benchmark for task-specific model building on heterogeneous, often multi-modal datasets. It contains 76 end-to-end tasks across 9 domains: sequence, single-cell, structure, network biology, chemical biology, perturbation dynamics, phenotype--disease, imaging, and text-integrated tasks. Each task packages primary-source data with hidden labels, held-out graders, and biology-aware metrics on a unified 0-to-100 scale; agents write runnable code, train models, and predict private test samples. BioXArena emphasizes realistic data interfaces: most tasks require agents to integrate multiple input sources, including 46 tasks that combine distinct modality families. We compare 18 harness--backbone configurations: general-purpose coding, biomedical, and ML agents plus our Minimal runner, all under the same controlled resource configuration. Codex with GPT-6 Astra achieves the highest task-level average score of 73.91 among 18 configurations, closely followed by Claude Code with Opus 5.5 (73.79). Matched-backbone comparisons show higher observed scores for mature coding systems than for Minimal; seven Minimal backends characterize model differences under a fixed runner. Beyond the main leaderboard, ablations, robustness checks, and scaling, runtime, and failure-mode analyses characterize how backbones, scaffolds, budgets, and domains affect BioML coding performance, exposing modeling and execution gaps across agent and LLM backbone families.

cs.CE↗
arXiv2026

Rent-Seeking Language Models: Reward Models Set the Prices, Frequent Audits Curb the Rent

Reward model overoptimization in RLHF is usually described as reward hacking. We recast it as rent-seeking: the policy, a firm, earns reward from a regulator, the proxy reward model, that true quality does not back. The lens turns overoptimization into measurable quantities, namely rent, a capture index and deadweight loss, and into testable predictions from the economics of regulation. In a fixed, version-controlled testbed with three model families, accuracy-matched proxy reward models and two gold reward models, RL turns 41% to 66% of its proxy gain into rent while gold reward plateaus. Rent follows each regulator's prices. Neither proxy verifies facts, but one pays for confident specifics and the other for displays of diligence, so RL policies learn to fabricate details about nonexistent entities (97% to 100% of probes) or to deflect factual questions (67% to 98%). Kinship between policy and regulator adds a fixed premium of about 0.1 standard deviations but gives the policy no extra leverage under optimization. At a fixed expected fine, frequent mild audits beat rare severe ones, and random audits are as efficient per audit as targeted ones. Bringing the bunching estimator of tax economics to RL training, we also find a first sign that policies learn to stay just below a targeted audit's threshold.

cs.LG↗
arXiv2026

BioDyad: Synchronize Biomedical Discovery and Machine Learning Engineering

Agentic biomedical machine learning (ML) draws on complementary advances in biomedical evidence acquisition and executable program search. Existing systems connect aspects of these capabilities, but coordinating them throughout program search remains challenging. New evidence must guide candidate construction, execution outcomes must inform subsequent discovery and reuse, and validation demands must fit the search budget. We introduce BioDyad, which couples biomedical discovery and ML engineering through two hierarchies within Monte Carlo graph search. Its scientific hierarchy combines prior biomedical guidance with iterative discovery, then links biomedical plans to execution outcomes in memory for reuse across candidates. Its engineering hierarchy moves candidate programs from smoke execution, through train/validation evaluation, to full-data retraining. We evaluate BioDyad on the 76-task BioXArena benchmark under a two-hour per-task budget with three matched LLM backends. It achieves the highest penalized all-task score and task success rate among four agent methods and a one-shot baseline under each backend. These results support coordinating biomedical discovery and ML engineering to integrate external knowledge into executable programs across heterogeneous biomedical tasks.

cs.AI↗
arXiv2026

Improved Variance Estimation in Homoskedastic Nonparametric Random-Design Regression via a Two-Scale Approach

We study estimation of a constant conditional variance $σ^2$ in nonparametric regression with a $d$-dimensional random design. This is an important problem, and similar questions arise in causal inference. The regression function is $β_b$-Hölder smooth, the design density is $β_g$-Hölder smooth and bounded above and away from zero, and we consider the nonparametric regime $β_b>1$ and $d>4β_b$. Set $β_g^\star=β_b(1-4β_b/d)/\{1+2β_b/d+8(β_b/d)^2\}$. We give an estimator whose mean squared error is upper bounded by $Cn^{-4(β_b+1)/(d+4)}$ in the low-regularity regime when $0<β_g\leqβ_g^\star$. The low-regularity branch is based on a new two-scale construction: the covariate space is partitioned into cells, the local polynomial trend is projected out within each suitable cell, and the squared normalized contrast from one eligible close pair per cell is averaged across cells. In the high regularity regime when $β_g>β_g^\star$, a higher-order influence function estimator of Robins, Li, Tchetgen Tchetgen, and van der Vaart (2008) provides the rate $Cn^{-8β_b/(d+4β_b)}$. We also give an all-pairs ridge extension, which achieves the same two-scale rate, and evaluate the methods alongside a range of existing estimators in simulations.

math.ST↗
arXiv2026

MedKGent: A Large Language Model Agent Framework for Constructing Temporally Evolving Medical Knowledge Graph

The rapid expansion of medical literature challenges the scalable structuring of domain knowledge. Knowledge Graphs (KGs) offer a solution, yet current construction methods lack generalizability and ignore the temporal dynamics of evolving knowledge. To address this, we introduce MedKGent, a Large Language Model (LLM) agent framework for building temporally evolving medical KGs. Using over 10 million PubMed abstracts from 1975 to 2023, MedKGent incrementally constructs a KG daily via two specialized agents. The Extractor Agent identifies knowledge triples and assigns confidence scores, while the Constructor Agent integrates these triples into a temporal graph, reinforcing recurring knowledge and resolving conflicts. The resulting KG contains 156,275 entities and 2,971,384 triples, making it, to our knowledge, the largest LLM-derived medical KG to date. Automated and expert assessments showed triple-validity rates approaching 90%. In downstream evaluations, MedKGent-KG significantly improved retrieval-augmented generation for five LLMs across seven medical question-answering benchmarks. Together, these results position MedKGent as a scalable and temporally aware infrastructure for medical knowledge representation and literature-grounded AI research.

cs.CL↗
arXiv2026

Remedying Coarsening-Based GNN Training under Heterophily via Adaptive Complementary Enhancement

Coarsening-based training for graph neural networks (GNNs), i.e.\ training on coarsened graphs rather than the original large ones, has become a promising direction for scaling GNNs to massive graphs. However, prior work has been evaluated almost exclusively on \textit{homophilic} graphs, leaving the more challenging \textit{heterophilic} settings underexplored. We show, both empirically and theoretically, that existing coarsening-based training methods suffer significant performance degradation on heterophilic graphs due to inevitable loss of graph information during coarsening. To address this, we propose {\bf A}daptive {\bf C}omplementary {\bf E}nhancement, a plug-and-play, model-agnostic strategy that reintegrates the information discarded in coarsening: ACE learns a projector for re-constructing original node features and applies \textit{anisotropic structural regularization} to embed local heterophily. We further adopt \textit{homoscedastic uncertainty weighting} to adaptively balance the combined training objective of primary coarsened-graph training loss and full-graph auxiliary loss with augmented node features re-constructed by the heterophily-aware projector. Extensive experiments show that ACE drives consistent gains on heterophilic benchmarks while preserving competitive results on homophilic graphs with minimal computational overhead. Code is available at the GitHub repository: https://github.com/vasile-paskardlgm/ACE.

cs.LG↗
arXiv2026

CSCO: A Backside-PDN-Aware Clock-Signal Co-Optimization Framework for Improved PPA

Backside power delivery networks (BSPDN) have emerged as a promising technology for advanced logic nodes to address IR-drop and PPA challenges. While BSPDN introduces additional routing resources on the backside, these resources are limited and must be carefully partitioned between clock and signal nets, creating a critical resource allocation tradeoff. Prior work either moves only the clock network or assumes a fixed clock tree and optimizes only signal nets, failing to explore the tradeoff space of backside resource allocation. Moreover, lacking frontside power-ground shielding, BSPDN introduces severe signal integrity (SI) degradation. We propose CSCO, a data-driven BSPDN-aware co-optimization framework that jointly allocates limited backside resources between clock and signal nets across frontside/backside layers. CSCO employs efficient search strategies to identify critical nets for backside routing without repeated evaluation, navigating the clock-signal allocation tradeoff to balance IR-drop, routing congestion, and PPA. The framework also leverages backside routing to mitigate coupling noise and crosstalk-induced SI issues. Experiments demonstrate improved WNS/TNS, frequency, and SI robustness without additional shielding overhead.

cs.CE↗
arXiv2026

FasterRW: An Efficient Prior-Assisted Random Walk Solver for Thermal Analysis

Thermal simulation is becoming increasingly critical in modern IC design and manufacturing. Feynman--Kac random walk methods can estimate localized temperatures without solving the global field, but in practical settings without Dirichlet boundaries they require very long paths to control the truncation residual. This paper presents FasterRW, a continuation-value framework for accelerating Feynman--Kac thermal random walks. We prove a stopping-time residual identity showing that the expected tail after truncation equals the current attenuation weight multiplied by the temperature at the stopping location. This identity enables a prior-corrected single-point estimator: a cheap, noisy prior replaces the unknown continuation value, making the truncation bias scale with prior error rather than absolute temperature and yielding an error-controlled threshold-selection rule. For multi-point queries, FasterRW further converts cross-query pass-through tails into pseudo-samples and refines the resulting observations in prior-error coordinates through kriging. Compared with prior state-of-the-art methods, FasterRW achieves up to $28\times$ acceleration while maintaining accuracy. Our code is open-sourced at https://github.com/ShiningSord/FastRW.

cs.LG↗
arXiv2026

Consistent Distributed Ranking of Generative Models via Kernel Distances

Ranking generative models based on the fidelity and diversity of their outputs is required to identify the best generator in a group of candidate generative AI models. To rank a group of models in a conventional centralized setting, a standard score is commonly evaluated for each involved model. The selection and design of reference-based evaluation scores have been extensively studied in centralized settings, where the reference samples are drawn from a single probability distribution. However, in practical scenarios including distributed learning contexts, reference samples are distributed across multiple clients, each potentially with a heterogeneous data distribution. In this work, we investigate the ranking of generative models in such distributed settings with heterogeneous data distributions across clients. We focus on the widely used family of kernel distance (KD) evaluation metrics. We prove that, for every kernel function, ranking models by the averaged KD scores of individual clients yields the same ordering as a centralized KD evaluation using the combined reference data from all the clients. We further extend our analysis to other popular metrics, including the Fréchet Distance (FD), for which the individual client scores could be insufficient for accurate model ranking. We present the numerical results of several experiments on standard image datasets and generative models to validate our theoretical findings regarding distributed ranking using various evaluation scores.

cs.LG↗
arXiv2026

OLion: Approaching the Hadamard Ideal by Intersecting Spectral and $\ell_{\infty}$ Implicit Biases

Many optimizers can be interpreted as steepest-descent methods under norm-induced geometries, and thus inherit corresponding implicit biases. We introduce \nameA{} (\fullname{}), which combines spectral control from orthogonalized update directions with $\ell_\infty$-style coordinate control from sign updates. \nameA{} forms a Lion-style momentum direction, approximately orthogonalizes it via a few Newton--Schulz iterations, and then applies an entrywise sign, providing an efficient approximation to taking a maximal step over the intersection of the spectral and $\ell_\infty$ constraint sets (a scaled Hadamard-like set for matrix parameters). Despite the strong nonlinearity of orthogonalization and sign, we prove convergence under a mild, empirically verified diagonal-isotropy assumption. Across large-scale language and vision training, including GPT-2 and Llama pretraining, SiT image pretraining, and supervised fine-tuning, \nameA{} matches or outperforms AdamW and Muon under comparable tuning while using only momentum-level optimizer state, and it mitigates optimizer mismatch when fine-tuning AdamW-pretrained checkpoints.

cs.LG↗
arXiv2026

RoboBrain 2.5: Depth in Sight, Time in Mind

We introduce RoboBrain 2.5, a next-generation embodied AI foundation model that advances general perception, spatial reasoning, and temporal modeling through extensive training on high-quality spatiotemporal supervision. Building upon its predecessor, RoboBrain 2.5 introduces two major capability upgrades. Specifically, it unlocks Precise 3D Spatial Reasoning by shifting from 2D pixel-relative grounding to depth-aware coordinate prediction and absolute metric constraint comprehension, generating complete 3D manipulation traces as ordered keypoint sequences under physical constraints. Complementing this spatial precision, the model establishes Dense Temporal Value Estimation that provides dense, step-aware progress prediction and execution state understanding across varying viewpoints, producing stable feedback signals for downstream learning. Together, these upgrades extend the framework toward more physically grounded and execution-aware embodied intelligence for complex, fine-grained manipulation. The code and checkpoints are available at project website: https://superrobobrain.github.io

cs.RO↗
arXiv2025

Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation

The primary obstacle for applying reinforcement learning (RL) to real-world robotics is the design of effective reward functions. While recently learning-based Process Reward Models (PRMs) are a promising direction, they are often hindered by two fundamental limitations: their reward models lack step-aware understanding and rely on single-view perception, leading to unreliable assessments of fine-grained manipulation progress; and their reward shaping procedures are theoretically unsound, often inducing a semantic trap that misguides policy optimization. To address these, we introduce Dopamine-Reward, a novel reward modeling method for learning a general-purpose, step-aware process reward model from multi-view inputs. At its core is our General Reward Model (GRM), trained on a vast 3,400+ hour dataset, which leverages Step-wise Reward Discretization for structural understanding and Multi-Perspective Reward Fusion to overcome perceptual limitations. Building upon Dopamine-Reward, we propose Dopamine-RL, a robust policy learning framework that employs a theoretically-sound Policy-Invariant Reward Shaping method, which enables the agent to leverage dense rewards for efficient self-improvement without altering the optimal policy, thereby fundamentally avoiding the semantic trap. Extensive experiments across diverse simulated and real-world tasks validate our approach. GRM achieves state-of-the-art accuracy in reward assessment, and Dopamine-RL built on GRM significantly improves policy learning efficiency. For instance, after GRM is adapted to a new task in a one-shot manner from a single expert trajectory, the resulting reward model enables Dopamine-RL to improve the policy from near-zero to 95% success with only 150 online rollouts (approximately 1 hour of real robot interaction), while retaining strong generalization across tasks. Project website: https://robo-dopamine.github.io

cs.RO↗
arXiv2025

Beyond Profile: From Surface-Level Facts to Deep Persona Simulation in LLMs

Previous approaches to persona simulation large language models (LLMs) have typically relied on learning basic biographical information, or using limited role-play dialogue datasets to capture a character's responses. However, a holistic representation of an individual goes beyond surface-level facts or conversations to deeper thoughts and thinking. In this work, we introduce CharacterBot, a model designed to replicate both the linguistic patterns and distinctive thought patterns as manifested in the textual works of a character. Using Lu Xun, a renowned Chinese writer as a case study, we propose four training tasks derived from his 17 essay collections. These include a pre-training task focused on mastering external linguistic structures and knowledge, as well as three fine-tuning tasks: multiple-choice question answering, generative question answering, and style transfer, each aligning the LLM with Lu Xun's internal ideation and writing style. To optimize learning across these tasks, we introduce a CharLoRA parameter updating mechanism, where a general linguistic style expert collaborates with other task-specific experts to better study both the language style and the understanding of deeper thoughts. We evaluate CharacterBot on three tasks for linguistic accuracy and opinion comprehension, demonstrating that it significantly outperforms the baselines on our adapted metrics. We hope this work inspires future research on deep character persona simulation LLMs while considering the importance of ethical standards.

cs.CL↗
arXiv2025

Can A Society of Generative Agents Simulate Human Behavior and Inform Public Health Policy? A Case Study on Vaccine Hesitancy

Can we simulate a sandbox society with generative agents to model human behavior, thereby reducing the over-reliance on real human trials for assessing public policies? In this work, we investigate the feasibility of simulating health-related decision-making, using vaccine hesitancy, defined as the delay in acceptance or refusal of vaccines despite the availability of vaccination services (MacDonald, 2015), as a case study. To this end, we introduce the VacSim framework with 100 generative agents powered by Large Language Models (LLMs). VacSim simulates vaccine policy outcomes with the following steps: 1) instantiate a population of agents with demographics based on census data; 2) connect the agents via a social network and model vaccine attitudes as a function of social dynamics and disease-related information; 3) design and evaluate various public health interventions aimed at mitigating vaccine hesitancy. To align with real-world results, we also introduce simulation warmup and attitude modulation to adjust agents' attitudes. We propose a series of evaluations to assess the reliability of various LLM simulations. Experiments indicate that models like Llama and Qwen can simulate aspects of human behavior but also highlight real-world alignment challenges, such as inconsistent responses with demographic profiles. This early exploration of LLM-driven simulations is not meant to serve as definitive policy guidance; instead, it serves as a call for action to examine social simulation for policy development.

cs.MA↗
arXiv2025

Test-Time Scaling with Reflective Generative Model

We introduce our first reflective generative model MetaStone-S1, which obtains OpenAI o3-mini's performance via the new Reflective Generative Form. The new form focuses on high-quality reasoning trajectory selection and contains two novelties: 1) A unified interface for policy and process reward model: we share the backbone network and use task-specific heads for reasoning trajectory predicting and scoring respectively, introducing only 53M extra parameters for trajectory scoring. 2) Eliminating the reliance on process-level annotation: we provide a self-supervised process reward model, which can directly learn the high-quality reasoning trajectory selection from the outcome reward. Equipped with the reflective generative form, MetaStone-S1 is naturally suitable for test-time scaling, and we provide three reasoning effort modes (low, medium, and high) based on the controllable thinking length. Experiments demonstrate that our MetaStone-S1 achieves comparable performance to OpenAI o3-mini's series with only 32B parameter size. To support the research community, we have open-sourced MetaStone-S1 at https://github.com/MetaStone-AI/MetaStone-S1.

cs.LG↗