Searcharxiv⌕ Search

arXiv subjects

Xinyi He

Publications and source records attributed to Xinyi He.

At least 19 recordsLinked to original sources

Probability of Success under Population Shift for Phase III Trial Planning

Probability of success (PoS) is widely used to inform go/no-go decisions for phase III trials based on phase II evidence. Conventional PoS calculations, however, implicitly assume that baseline covariate distributions are identical across phases. When covariate shift exists between phase II and phase III populations, unadjusted PoS may be biased and lead to miscalibrated decision-making. We propose a Bayesian framework for PoS evaluation that explicitly accounts for covariate shift between phase II and III population prior to phase III initiation. The method adopts a transportability perspective, adjusting phase II evidence to reflect an anticipated phase III target population using available target covariate information. The framework accommodates both individual-level and aggregate-level phase III covariate data. Simulation studies demonstrate that, compared with the conventional approach, the proposed method improves PoS accuracy and decision calibration when covariate shift is present, while maintaining comparable performance when populations are aligned. We illustrate the proposed methodology using clinical trial data from systemic lupus erythematosus.

stat.ME↗

EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents

Continual learning requires models to accumulate past experience and transfer it to new tasks. Despite substantial research on continual learning, existing work remains largely confined to non-embodied settings, without extending to embodied environments. To address this gap, we introduce EmbodiedCLBench, the first benchmark in embodied simulation for evaluating the continual learning capability of self-evolving agents. With compositional generalization as the core evaluation principle, our benchmark evaluates whether agents can first learn from basic tasks and then recombine the acquired skills to solve complex, unseen advanced tasks. Based on EmbodiedCLBench, we conduct extensive experiments across multiple agent harnesses and learning methods, revealing consistent limitations and distinctive behaviors. Our results show that while advanced tasks can be hardly solved under zero-shot setting, the experience from basic tasks enables consistent improvement. Such benefit from the experience remains stable even if the task complexity increases. However, this improvement does not scale accordingly with the number of learning samples, and current agents cannot extract meaningful information from additional experience. Collectively, our findings highlight continual learning as a fundamental yet underexplored capability, and our benchmark offers valuable resources for designing effective self-evolving agents in embodied environments. Our code and dataset are available at https://pku-value-lab.github.io/EmbodiedCLBench-homepage.

cs.AI↗

Adaptively Combining Randomized and External Control Data Using a Mixture Prior in the Presence of Heterogeneity

When resources are limited in a randomized trial, a common strategy is to augment data from the control arm with external controls. We propose a hybrid group sequential Bayesian design that uses this approach to compare time-to-event distributions. The design allows between-treatment effects to differ between patient subgroups, and adaptively combines subgroups that have similar hazard functions. The aim is to reduce the control arm sample size without compromising the validity of comparative inferences due to systematic trial-versus-external data differences. The model borrows external control data dynamically by using a self-adapting mixture prior, which is a weighted average of a non-informative prior and an informative prior constructed from the external controls. The amount of borrowing is proportional to the agreement between the randomized and external controls. At each interim analysis, the design compares the treatments in each subgroup, and if inferiority or superiority is concluded stops accrual for that subgroup. Simulations show that the proposed design outperforms other methods for incorporating external control data in a randomized trial, and improves efficiency.

stat.ME↗

Kimi K3: Open Frontier Intelligence

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence.

cs.CL↗

SWE-Perf: Can Language Models Optimize Code Performance on Real-World Repositories?

Code performance optimization is paramount in real-world software engineering and critical for production-level systems. While Large Language Models (LLMs) have demonstrated impressive capabilities in code generation and bug fixing, their proficiency in enhancing code performance at the repository level remains largely unexplored. To address this gap, we introduce SWE-Perf, the first benchmark specifically designed to systematically evaluate LLMs on code performance optimization tasks within authentic repository contexts. SWE-Perf comprises 140 carefully curated instances, each derived from performance-improving pull requests from popular GitHub repositories. Each benchmark instance includes the relevant codebase, target functions, performance-related tests, expert-authored patches, and executable environments. Through a comprehensive evaluation of representative methods that span file-level and repo-level approaches (e.g., Agentless and OpenHands), we reveal a substantial capability gap between existing LLMs and expert-level optimization performance, highlighting critical research opportunities in this emerging field.

cs.SE↗

Automatic Cognitive Task Generation for In-Situ Evaluation of Embodied Agents

As general intelligent agents are poised for widespread deployment in diverse households, evaluation tailored to each unique unseen 3D environment has become a critical prerequisite. However, existing benchmarks suffer from severe data contamination and a lack of scene specificity, inadequate for assessing agent capabilities in unseen settings. To address this, we propose a dynamic in-situ task generation method for unseen environments inspired by human cognition. We define tasks through a structured graph representation and construct a two-stage interaction-evolution task generation system for embodied agents (TEA). In the interaction stage, the agent actively interacts with the environment, creating a loop between task execution and generation that allows for continuous task generation. In the evolution stage, task graph modeling allows us to recombine and reuse existing tasks to generate new ones without external data. Experiments across 10 unseen scenes demonstrate that TEA automatically generated 87,876 tasks in two cycles, which human verification confirmed to be physically reasonable and encompassing essential daily cognitive capabilities. Benchmarking SOTA models against humans on our in-situ tasks reveals that models, despite excelling on public benchmarks, perform surprisingly poorly on basic perception tasks, severely lack 3D interaction awareness and show high sensitivity to task types in reasoning. These sobering findings highlight the necessity of in-situ evaluation before deploying agents into real-world human environments.

cs.AI↗

TongSIM: A General Platform for Simulating Intelligent Machines

As artificial intelligence (AI) rapidly advances, especially in multimodal large language models (MLLMs), research focus is shifting from single-modality text processing to the more complex domains of multimodal and embodied AI. Embodied intelligence focuses on training agents within realistic simulated environments, leveraging physical interaction and action feedback rather than conventionally labeled datasets. Yet, most existing simulation platforms remain narrowly designed, each tailored to specific tasks. A versatile, general-purpose training environment that can support everything from low-level embodied navigation to high-level composite activities, such as multi-agent social simulation and human-AI collaboration, remains largely unavailable. To bridge this gap, we introduce TongSIM, a high-fidelity, general-purpose platform for training and evaluating embodied agents. TongSIM offers practical advantages by providing over 100 diverse, multi-room indoor scenarios as well as an open-ended, interaction-rich outdoor town simulation, ensuring broad applicability across research needs. Its comprehensive evaluation framework and benchmarks enable precise assessment of agent capabilities, such as perception, cognition, decision-making, human-robot cooperation, and spatial and social reasoning. With features like customized scenes, task-adaptive fidelity, diverse agent types, and dynamic environmental simulation, TongSIM delivers flexibility and scalability for researchers, serving as a unified platform that accelerates training, evaluation, and advancement toward general embodied intelligence.

cs.AI↗

Modeling Atmospheric Ion Escape from Kepler-1649 b and c over Time

Rocky planets orbiting M-dwarf stars are prime targets for atmospheric characterization, yet their long-term evolution under intense stellar winds and high-energy radiation remains poorly constrained. The Kepler-1649 system, hosting two terrestrial exoplanets orbiting an M5V star, provides a valuable laboratory for studying atmospheric evolution in the extreme environments typical of M-dwarf systems. In this Letter we show that both planets could have retained atmospheres over gigayear timescales. Using a multi-species magnetohydrodynamic model, we simulate atmospheric ion escape driven by stellar winds and extreme ultraviolet radiation from 0.8 to 4.0 Gyr. The results reveal a clear decline in total ion escape rates with stellar age, as captured by a nonparametric LOWESS regression, with O$^{+}$ comprising 98.3%-99.9% of the total loss. Escape rates at 4.0 Gyr are two to three orders of magnitude lower than during early epochs. At 0.8 Gyr, planet b exhibits 3.79$\times$ higher O$^{+}$ escape rates than planet c, whereas by 4.0 Gyr its O$^{+}$ escape rate becomes 39.5$\times$ lower. This reversal arises from a transition to sub-magnetosonic star-planet interactions, where the fast magnetosonic Mach number, $M_f$, falls below unity. Despite substantial early atmospheric erosion, both planets may have retained significant atmospheres, suggesting potential long-term habitability. These findings offer predictive insight into atmospheric retention in the Kepler-1649 system and inform future JWST observations of similar M-dwarf terrestrial exoplanets aimed at refining habitability assessments.

astro-ph.EP↗

Evaluating Multimodal Large Language Models with Daily Composite Tasks in Home Environments

A key feature differentiating artificial general intelligence (AGI) from traditional AI is that AGI can perform composite tasks that require a wide range of capabilities. Although embodied agents powered by multimodal large language models (MLLMs) offer rich perceptual and interactive capabilities, it remains largely unexplored whether they can solve composite tasks. In the current work, we designed a set of composite tasks inspired by common daily activities observed in early childhood development. Within a dynamic and simulated home environment, these tasks span three core domains: object understanding, spatial intelligence, and social activity. We evaluated 17 leading proprietary and open-source MLLMs on these tasks. The results consistently showed poor performance across all three domains, indicating a substantial gap between current capabilities and general intelligence requirements. Together, our tasks offer a preliminary framework for evaluating the general capabilities of embodied agents, marking an early but significant step toward the development of embodied MLLMs and their real-world deployment.

cs.AI↗

TableLoRA: Low-rank Adaptation on Table Structure Understanding for Large Language Models

Tabular data are crucial in many fields and their understanding by large language models (LLMs) under high parameter efficiency paradigm is important. However, directly applying parameter-efficient fine-tuning (PEFT) techniques to tabular tasks presents significant challenges, particularly in terms of better table serialization and the representation of two-dimensional structured information within a one-dimensional sequence. To address this, we propose TableLoRA, a module designed to improve LLMs' understanding of table structure during PEFT. It incorporates special tokens for serializing tables with special token encoder and uses 2D LoRA to encode low-rank information on cell positions. Experiments on four tabular-related datasets demonstrate that TableLoRA consistently outperforms vanilla LoRA and surpasses various table encoding methods tested in control experiments. These findings reveal that TableLoRA, as a table-specific LoRA, enhances the ability of LLMs to process tabular data effectively, especially in low-parameter settings, demonstrating its potential as a robust solution for handling table-related tasks.

cs.CL↗

Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization

Large Language Models (LLMs) generate functionally correct solutions but often fall short in code efficiency, a critical bottleneck for real-world deployment. In this paper, we introduce a novel test-time iterative optimization framework to address this, employing a closed-loop system where LLMs iteratively refine code based on empirical performance feedback from an execution sandbox. We explore three training strategies: Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Group Relative Policy Optimization (GRPO). Experiments on our Venus dataset and the APPS benchmark show that SFT and DPO rapidly saturate in efficiency gains. In contrast, GRPO, using reinforcement learning (RL) with execution feedback, continuously optimizes code performance, significantly boosting both pass@1 (from 47% to 62%) and the likelihood of outperforming human submissions in efficiency (from 31% to 45%). Our work demonstrates effective test-time code efficiency improvement and critically reveals the power of RL in teaching LLMs to truly self-improve code efficiency.

cs.SE↗

Diffusion Suction Grasping with Large-Scale Parcel Dataset

While recent advances in object suction grasping have shown remarkable progress, significant challenges persist particularly in cluttered and complex parcel handling scenarios. Two fundamental limitations hinder current approaches: (1) the lack of a comprehensive suction grasp dataset tailored for parcel manipulation tasks, and (2) insufficient adaptability to diverse object characteristics including size variations, geometric complexity, and textural diversity. To address these challenges, we present Parcel-Suction-Dataset, a large-scale synthetic dataset containing 25 thousand cluttered scenes with 410 million precision-annotated suction grasp poses. This dataset is generated through our novel geometric sampling algorithm that enables efficient generation of optimal suction grasps incorporating both physical constraints and material properties. We further propose Diffusion-Suction, an innovative framework that reformulates suction grasp prediction as a conditional generation task through denoising diffusion probabilistic models. Our method iteratively refines random noise into suction grasp score maps through visual-conditioned guidance from point cloud observations, effectively learning spatial point-wise affordances from our synthetic dataset. Extensive experiments demonstrate that the simple yet efficient Diffusion-Suction achieves new state-of-the-art performance compared to previous models on both Parcel-Suction-Dataset and the public SuctionNet-1Billion benchmark.

cs.CV↗

CoCoST: Automatic Complex Code Generation with Online Searching and Correctness Testing

Large Language Models have revolutionized code generation ability by converting natural language descriptions into executable code. However, generating complex code within real-world scenarios remains challenging due to intricate structures, subtle bugs, understanding of advanced data types, and lack of supplementary contents. To address these challenges, we introduce the CoCoST framework, which enhances complex code generation by online searching for more information with planned queries and correctness testing for code refinement. Moreover, CoCoST serializes the complex inputs and outputs to improve comprehension and generates test cases to ensure the adaptability for real-world applications. CoCoST is validated through rigorous experiments on the DS-1000 and ClassEval datasets. Experimental results show that CoCoST substantially improves the quality of complex code generation, highlighting its potential to enhance the practicality of LLMs in generating complex code.

cs.SE↗

TAP4LLM: Table Provider on Sampling, Augmenting, and Packing Semi-structured Data for Large Language Model Reasoning

Table reasoning tasks have shown remarkable progress with the development of large language models (LLMs), which involve interpreting and drawing conclusions from tabular data based on natural language (NL) questions. Existing solutions mainly tested on smaller tables face scalability issues and struggle with complex queries due to incomplete or dispersed data across different table sections. To alleviate these challenges, we propose TAP4LLM as a versatile pre-processor suite for leveraging LLMs in table-based tasks effectively. It covers several distinct components: (1) table sampling to decompose large tables into manageable sub-tables based on query semantics, (2) table augmentation to enhance tables with additional knowledge from external sources or models, and (3) table packing & serialization to convert tables into various formats suitable for LLMs' understanding. In each module, we design and compare several common methods under various usage scenarios, aiming to shed light on the best practices for leveraging LLMs for table-reasoning tasks. Our experiments show that our method improves LLMs' reasoning capabilities in various tabular tasks and enhances the interaction between LLMs and tabular data by employing effective pre-processing.

cs.CL↗

The association of domain-specific physical activity and sedentary activity with stroke: A prospective cohort study

Background The incidence of stroke places a heavy burden on both society and individuals. Activity is closely related to cardiovascular health. This study aimed to investigate the relationship between the varying domains of PA, like occupation-related Physical Activity (OPA), transportation-related Physical Activity (TPA), leisure-time Physical Activity (LTPA), and Sedentary Activity (SA) with stroke. Methods Our analysis included 30,400 participants aged 20+ years from 2007 to 2018 National Health and Nutrition Examination Survey (NHANES). Stroke was identified based on the participant's self-reported diagnoses from previous medical consultations, and PA and SA were self-reported. Multivariable logistic and restricted cubic spline models were used to assess the associations. Results Participants achieving PA guidelines (performing PA more than 150 min/week) were 35.7% less likely to have a stroke based on both the total PA (odds ratio [OR] 0.643, 95% confidence interval [CI] 0.523-0.790) and LTPA (OR 0.643, 95% CI 0.514-0.805), while OPA or TPA did not demonstrate lower stroke risk. Furthermore, participants with less than 7.5 h/day SA levels were 21.6% (OR 0.784, 95% CI 0.665-0.925) less likely to have a stroke. The intensities of total PA and LTPA exhibited nonlinear U-shaped associations with stroke risk. In contrast, those of OPA and TPA showed negative linear associations, while SA intensities were positively linearly correlated with stroke risk. Conclusions LTPA, but not OPA or TPA, was associated with a lower risk of stroke at any amount, suggesting that significant cardiovascular health would benefit from increased PA. Additionally, the positive association between SA and stroke indicated that prolonged sitting was detrimental to cardiovascular health. Overall, increased PA within a reasonable range reduces the risk of stroke, while increased SA elevates it.

physics.med-ph↗

Discovery of self-assembled Ru/Si heterostructures with unique periodic nanostripe patterns boosting hydrogen evolution

Two-dimensional (2D) heterostructuring is a versatile methodology for designing nanoarchitecture catalytic systems that allow for reconstruction and modulation of interfaces and electronic structures. However, catalysts with such structures are extremely scarce due to limited synthetic strategies. Here, we report a highly ordered 2D Ru/Si nano-heterostructures (RSHS) by acid etching of the LaRuSi electride. RSHS shows a superior electrocatalytic activity for hydrogen evolution with an overpotential of 14 mV at 10 mA/cm2 in alkaline media. Both experimental analysis and first-principles calculations demonstrate that the electronic states of Ru can be tuned by strong interactions of the interfacial Ru-Si, leading to an optimized hydrogen adsorption energy. Moreover, due to the synergistic effect of Ru and Si, the energy barrier of water dissociation is significantly reduced. The unique nanostripe structure with abundant interfaces in RSHS will provide a paradigm for construction of efficient catalysts with tunable electronic states and dual active sites.

cond-mat.mtrl-sci↗

Text2Analysis: A Benchmark of Table Question Answering with Advanced Data Analysis and Unclear Queries

Tabular data analysis is crucial in various fields, and large language models show promise in this area. However, current research mostly focuses on rudimentary tasks like Text2SQL and TableQA, neglecting advanced analysis like forecasting and chart generation. To address this gap, we developed the Text2Analysis benchmark, incorporating advanced analysis tasks that go beyond the SQL-compatible operations and require more in-depth analysis. We also develop five innovative and effective annotation methods, harnessing the capabilities of large language models to enhance data quality and quantity. Additionally, we include unclear queries that resemble real-world user questions to test how well models can understand and tackle such challenges. Finally, we collect 2249 query-result pairs with 347 tables. We evaluate five state-of-the-art models using three different metrics and the results show that our benchmark presents introduces considerable challenge in the field of tabular data analysis, paving the way for more advanced research opportunities.

cs.CL↗

GPT4Graph: Can Large Language Models Understand Graph Structured Data ? An Empirical Evaluation and Benchmarking

Large language models~(LLM) like ChatGPT have become indispensable to artificial general intelligence~(AGI), demonstrating excellent performance in various natural language processing tasks. In the real world, graph data is ubiquitous and an essential part of AGI and prevails in domains like social network analysis, bioinformatics and recommender systems. The training corpus of large language models often includes some algorithmic components, which allows them to achieve certain effects on some graph data-related problems. However, there is still little research on their performance on a broader range of graph-structured data. In this study, we conduct an extensive investigation to assess the proficiency of LLMs in comprehending graph data, employing a diverse range of structural and semantic-related tasks. Our analysis encompasses 10 distinct tasks that evaluate the LLMs' capabilities in graph understanding. Through our study, we not only uncover the current limitations of language models in comprehending graph structures and performing associated reasoning tasks but also emphasize the necessity for further advancements and novel approaches to enhance their graph processing capabilities. Our findings contribute valuable insights towards bridging the gap between language models and graph understanding, paving the way for more effective graph mining and knowledge extraction.

cs.AI↗