Author name cluster

Yutong Wu

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

1 author row

AAAI Conference 2026 Conference Paper

StepFun-Formalizer: Unlocking the Autoformalization Potential of LLMs Through Knowledge-Reasoning Fusion

Yutong Wu
Di Huang
Ruosi Wan
Yue Peng
Shijie Shang
Chenrui Cao
Lei Qi
Rui Zhang

Autoformalization aims to translate natural-language mathematical statements into a formal language. While LLMs have accelerated progress in this area, existing methods still suffer from low accuracy. We identify two key abilities for effective autoformalization: comprehensive mastery of formal-language domain knowledge, and reasoning capability of natural language problem understanding and informal-formal alignment. Without the former, a model cannot identify the correct formal objects; without the latter, it struggles to interpret real-world contexts and map them precisely into formal expressions. To address these gaps, we introduce ThinkingF, a data synthesis and training pipeline that improves both abilities. First, we construct two datasets: one by distilling and selecting large-scale examples rich in formal knowledge, and another by generating informal-to-formal reasoning trajectories guided by expert-designed templates. We then apply SFT and RLVR with these datasets to further fuse and refine the two abilities. The resulting 7B and 32B models exhibit both comprehensive formal knowledge and strong informal-to-formal reasoning. Notably, StepFun-Formalizer-32B achieves SOTA BEq@1 scores of 40.5% on FormalMATH-Lite and 26.7% on ProverBench, surpassing all prior general-purpose and specialized models.

PDF Details DOI

YNIMG Journal 2026 Journal Article

System-Level Reconfiguration of the Aging Brain: Linking Dynamics, Morphology and Micro-architectures

Liming Fan
Youjun Li
Yutong Wu
Simeng An
Nan Yao
Qian Zhu
Yueye Zhao
Daqing Guo

Healthy aging involves complex neural reconfigurations across both structural and functional domains. While resting-state functional magnetic resonance imaging (rs-fMRI) has linked static functional connectivity alterations to aging, the whole-brain dynamics of functional activity and their covariance with structural changes remain poorly characterized. To address this gap, we integrated three data-driven approaches to profile functional dynamics in the aging brain and decode their association with structural atrophy. Using rs-fMRI data from 252 participants-145 young adults (22.7 ± 3.4 years) and 107 older adults (68.7 ± 6.5 years)-we made several key observations. First, normalized Shannon entropy revealed a significant reduction in spatiotemporal complexity among older individuals. Second, phase synchronization analysis of BOLD signals indicated enhanced global integration and metastability in older adults, particularly within the dorsal attention (DAN), ventral attention (VAN), and frontoparietal networks (FPN). Third, temporal asymmetry analysis demonstrated increased nonreversibility and a heightened functional hierarchy in the aging brain, again most evident in the FPN. Morphometric analyses confirmed widespread structural atrophy in older participants. Crucially, partial least squares (PLS) analysis uncovered significant covariance between morphometric patterns and dynamic functional metrics, underscoring a tight structure-dynamics coupling in aging. Furthermore, structural atrophy correlated significantly with variations in micro-architecture maps. Finally, we evaluated the behavioral relevance of these dynamics through correlations with cognitive performance. Our findings offer an integrative, multiscale perspective on neural decline in aging, emphasizing the interplay between dynamic functional reorganization and structural atrophy.

Details DOI

YNIMG Journal 2025 Journal Article

Brain dynamics alterations induced by partial sleep deprivation: An energy landscape study

Yutong Wu
Liming Fan
Wei Chen
Xing Su
Simeng An
Nan Yao
Qian Zhu
Zi-gang Huang

Partial sleep deprivation (PSD) alters neural activity of intrinsic brain networks involved in cognitive functions. However, the age-related time-varying properties of large-scale brain functional networks after PSD remain unknown. Our study applied energy landscape analysis to resting-state functional magnetic resonance imaging data to characterize the dominant brain activity patterns in 36 healthy young (19 females, 23. 53 ± 2. 36 years) and 33 healthy older (18 females, 68. 81 ± 2. 41 years) adults after full sleep (FS) and PSD. Dynamic properties of these patterns, including appearance probability, duration and transitions, were then calculated. Finally, a 105 steps numerical simulation was performed on each energy landscape. We found that the energy landscapes of the younger and older groups had similar hierarchical structures, including two major states and two minor states. The two major states showed complementary spontaneous activation patterns. But the PSD has altered the temporal evolution of these major brain states in younger participants, manifested by significantly higher appearance frequency of the major states and the direct transitions between major states than FS. These changes were not significant in older participants. Additionally, the weaker functional segregation between two modules assigned by two complementary major states was found during PSD than FS in young group. We further demonstrated that such abnormal brain network functional coordination was associated with the atypical brain dynamics and behaviors. These findings suggested a low-dimensional and restricted dynamic landscape of brain activity in young adults after PSD and provided new insight into understand the neural effects of PSD.

Details DOI

AAAI Conference 2025 Conference Paper

Functional Connectomes of Neural Networks

Tananun Songdechakraiwut
Yutong Wu

The human brain is a complex system, and understanding its mechanisms has been a long-standing challenge in neuroscience. The study of the functional connectome, which maps the functional connections between different brain regions, has provided valuable insights through various advanced analysis techniques developed over the years. Similarly, neural networks, inspired by the brain's architecture, have achieved notable success in diverse applications but are often noted for their lack of interpretability. In this paper, we propose a novel approach that bridges neural networks and human brain functions by leveraging brain-inspired techniques. Our approach, grounded in the insights from the functional connectome, offers scalable ways to characterize topology of large neural networks using stable statistical and machine learning techniques. Our empirical analysis demonstrates its capability to enhance the interpretability of neural networks, providing a deeper understanding of their underlying mechanisms.

PDF Details DOI

AIJ Journal 2025 Journal Article

IID prophet inequality with a single data point

Yilong Feng
Bo Li
Haolong Li
Xiaowei Wu
Yutong Wu

In this work, we study the single-choice prophet inequality problem, where a seller encounters a sequence of n online bids. These bids are modeled as independent and identically distributed (i. i. d.) random variables drawn from an unknown distribution. Upon the revelation of each bid's value, the seller must make an immediate and irrevocable decision on whether to accept the bid and sell the item to the bidder. The objective is to maximize the competitive ratio between the expected gain of the seller and that of the maximum bid. It is shown by Correa et al. [1] that when the distribution is unknown or only o ( n ) uniform samples from the distribution are given, the best an algorithm can do is 1 / e -competitive. In contrast, when the distribution is known [2], or when Ω ( n ) uniform samples are given [3], the optimal competitive ratio of 0. 7451 can be achieved. In this paper, we study the setting when the seller has access to a single point in the cumulative density function of the distribution, which can be learned from historical sales data. We investigate how effectively this data point can be used to design competitive algorithms. Motivated by the algorithm for the secretary problem, we propose the observe-and-accept algorithm that sets a threshold in the first phase using the data point and adopts the highest bid from the first phase as the threshold for the second phase. It can be viewed as a natural combination of the single-threshold algorithm for prophet inequality and the secretary problem algorithm. We show that our algorithm achieves a good competitive ratio for a wide range of data points, reaching up to 0. 6785-competitive as n → ∞ for certain data points. Additionally, we study an extension of the algorithm that utilizes more than two phases and show that the competitive ratio can be further improved to at least 0. 6862.

Details DOI

AAAI Conference 2025 Conference Paper

InverseCoder: Self-improving Instruction-Tuned Code LLMs with Inverse-Instruct

Yutong Wu
Di Huang
Wenxuan Shi
Wei Wang
Yewen Pu
Lingzhe Gao
Shihao Liu
Ziyuan Nan

Recent advancements in open-source code large language models (LLMs) have been driven by fine-tuning on the data generated from powerful closed-source LLMs, which are expensive to obtain. This paper explores whether it is possible to use a fine-tuned open-source model to generate additional data to augment its instruction-tuning dataset. We make two observations: (1) A code snippet can serve as the response to different instructions. (2) Instruction-tuned code LLMs perform better at translating code into instructions than the reverse. Based on these observations, we propose Inverse-Instruct, a data augmentation technique that uses a fine-tuned LLM to generate additional instructions of code responses from its own training dataset. The additional instruction-response pairs are added to the original dataset, and a stronger code LLM can be obtained by fine-tuning on the augmented dataset. We empirically validate Inverse-Instruct on a range of open-source code models (e.g. CodeLlama-Python and DeepSeek-Coder) and benchmarks (e.g., HumanEval(+), MBPP(+), DS-1000 and MultiPL-E), showing it consistently improves the base models.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

QiMeng-CodeV-R1: Reasoning-Enhanced Verilog Generation

Yaoyu Zhu
Di Huang
Hanqi Lyu
Xiaoyun Zhang
Chongxiao Li
Wenxuan Shi
Yutong Wu
Jianan Mu

Large language models (LLMs) trained via reinforcement learning with verifiable reward (RLVR) have achieved breakthroughs on tasks with explicit, automatable verification, such as software programming and mathematical problems. Extending RLVR to electronic design automation (EDA), especially automatically generating hardware description languages (HDLs) like Verilog from natural-language (NL) specifications, however, poses three key challenges: the lack of automated and accurate verification environments, the scarcity of high-quality NL-code pairs, and the prohibitive computation cost of RLVR. To this end, we introduce CodeV-R1, an RLVR framework for training Verilog generation LLMs. First, we develop a rule-based testbench generator that performs robust equivalence checking against golden references. Second, we propose a round-trip data synthesis method that pairs open-source Verilog snippets with LLM-generated NL descriptions, verifies code–NL–code consistency via the generated testbench, and filters out inequivalent examples to yield a high-quality dataset. Third, we employ a two-stage "distill-then-RL" training pipeline: distillation for the cold start of reasoning abilities, followed by adaptive DAPO, our novel RLVR algorithm that can reduce training cost by adaptively adjusting sampling rate. The resulting model, CodeV-R1-7B, achieves 68. 6 \% and 72. 9 \% pass@1 on VerilogEval v2 and RTLLM v1. 1, respectively, surpassing prior state-of-the-art by 12$\sim$20 \%, while even exceeding the performance of 671B DeepSeek-R1 on RTLLM. We have released our model, training code, and dataset to facilitate research in EDA and LLM communities.

PDF Details

TCS Journal 2024 Journal Article

Eliciting truthful reports with partial signals in repeated games

Yutong Wu
Ali Khodabakhsh
Bo Li
Evdokia Nikolova
Emmanouil Pountourakis

We consider a repeated game where a player self-reports her usage of a service and is charged a payment accordingly by a center. The center observes a partial signal, representing part of the player's true consumption, which is generated from a publicly known distribution. The player can report any value that does not contradict the signal and the center issues a payment based on the reported information. Such problems find application in net metering billing in the electricity market, where a customer's actual consumption of the electricity network is masked and complete verification is impractical. When the underlying true value is relatively constant, we propose a penalty mechanism that elicits truthful self-reports. Namely, besides charging the player the reported value, the mechanism charges a penalty proportional to her inconsistent reports. We show how fear of uncertainty in the future incentivizes the player to be truthful today. For Bernoulli distributions, we give the complete analysis and optimal strategies given any penalty. Since complete truthfulness is not possible for continuous distributions, we give approximate truthful results by a reduction from Bernoulli distributions. We also extend our mechanism to a multi-player cost-sharing setting and give equilibrium results.

Details DOI

AAAI Conference 2024 Conference Paper

Personalization as a Shortcut for Few-Shot Backdoor Attack against Text-to-Image Diffusion Models

Yihao Huang
Felix Juefei-Xu
Qing Guo
Jie Zhang
Yutong Wu
Ming Hu
Tianlin Li
Geguang Pu

Although recent personalization methods have democratized high-resolution image synthesis by enabling swift concept acquisition with minimal examples and lightweight computation, they also present an exploitable avenue for highly accessible backdoor attacks. This paper investigates a critical and unexplored aspect of text-to-image (T2I) diffusion models - their potential vulnerability to backdoor attacks via personalization. By studying the prompt processing of popular personalization methods (epitomized by Textual Inversion and DreamBooth), we have devised dedicated personalization-based backdoor attacks according to the different ways of dealing with unseen tokens and divide them into two families: nouveau-token and legacy-token backdoor attacks. In comparison to conventional backdoor attacks involving the fine-tuning of the entire text-to-image diffusion model, our proposed personalization-based backdoor attack method can facilitate more tailored, efficient, and few-shot attacks. Through comprehensive empirical study, we endorse the utilization of the nouveau-token backdoor attack due to its impressive effectiveness, stealthiness, and integrity, markedly outperforming the legacy-token backdoor attack.

PDF Details DOI