Yifei Li

Yifei Li, Pengyiang Liu, Yuhang Zang et al.

Multimodal agents in robotics, AR, and autonomous driving must reason about places and layouts from continuous egocentric streams, often using evidence outside the current view. Existing benchmarks either evaluate offline over full videos or target events rather than spatial structure. We introduce OVO-S-Bench, a fully human-annotated benchmark for streaming spatial intelligence, comprising 1,680 questions over 348 source videos. Annotation involves 12 trained annotators, each also serving as a blind cross-reviewer, across roughly 804 person-hours of multi-round quality assurance. Each question carries a query timestamp and an evidence interval, and at evaluation, the model sees only the prefix preceding the query. Questions span four levels of increasing abstraction: instantaneous egocentric perception, spatiotemporal context tracking, spatial simulation and reasoning, and allocentric mapping. Across 38 proprietary and open-source MLLMs, Gemini-3.1-Pro trails human experts by 27 points, 59.2 vs. 86.6, with allocentric mapping as the dominant bottleneck. Notably, streaming and spatially fine-tuned MLLMs underperform their own backbones. We further find that chain-of-thought reasoning amplifies spatial errors when ungrounded in the stream. By exposing these limitations, OVO-S-Bench establishes a demanding testbed for next-generation streaming spatial MLLMs.

Weizhu Bao, Yifei Li, Dongmin Wang

This work develops novel energy-stable parametric finite element methods (ES-PFEM) for the Willmore flow and curvature-dependent geometric gradient flows of surfaces in three dimensions. The key to achieving the energy stability lies in the use of two novel geometric identities: (i) a reformulated variational form of the normal velocity field, and (ii) incorporation of the temporal evolution of the mean curvature into the governing equations. These identities enable the derivation of a new variational formulation. By using the parametric finite element method, an implicit fully discrete scheme is subsequently developed, which maintains the energy dissipative property at the fully discrete level. Based on the ES-PFEM, comprehensive insights into the design of ES-PFEM for general curvature-dependent geometric gradient flows and a new understanding of mesh quality improvement in PFEM are provided. In particular, we develop the first PFEM for the Gauss curvature flow of surfaces. Furthermore, a tangential velocity control methodology is applied to improve the mesh quality and enhance the robustness of the proposed numerical method. Extensive numerical experiments confirm that the proposed method preserves energy dissipation properties and maintain good mesh quality in the surface evolution under the Willmore flow.

Martin Josifoski, Lars Klein, Maxime Peyrard et al.

Recent advances in artificial intelligence (AI) have produced highly capable and controllable systems. This creates unprecedented opportunities for structured reasoning as well as collaboration among multiple AI systems and humans. To fully realize this potential, it is essential to develop a principled way of designing and studying such structured interactions. For this purpose, we introduce the conceptual framework Flows. Flows are self-contained building blocks of computation, with an isolated state, communicating through a standardized message-based interface. This modular design simplifies the process of creating Flows by allowing them to be recursively composed into arbitrarily nested interactions and is inherently concurrency-friendly. Crucially, any interaction can be implemented using this framework, including prior work on AI-AI and human-AI interactions, prompt engineering schemes, and tool augmentation. We demonstrate the potential of Flows on competitive coding, a challenging task on which even GPT-4 struggles. Our results suggest that structured reasoning and collaboration substantially improve generalization, with AI-only Flows adding +21 and human-AI Flows adding +54 absolute points in terms of solve rate. To support rapid and rigorous research, we introduce the aiFlows library embodying Flows. The aiFlows library is available at https://github.com/epfl-dlab/aiflows. Data and Flows for reproducing our experiments are available at https://github.com/epfl-dlab/cc_flows.

Shengnan An, Yifei Li, Zeqi Lin et al. · pku

Recently the prompt-tuning paradigm has attracted significant attention. By only tuning continuous prompts with a frozen pre-trained language model (PLM), prompt-tuning takes a step towards deploying a shared frozen PLM to serve numerous downstream tasks. Although prompt-tuning shows good performance on certain natural language understanding (NLU) tasks, its effectiveness on natural language generation (NLG) tasks is still under-explored. In this paper, we argue that one of the factors hindering the development of prompt-tuning on NLG tasks is the unfamiliar inputs (i.e., inputs are linguistically different from the pretraining corpus). For example, our preliminary exploration reveals a large performance gap between prompt-tuning and fine-tuning when unfamiliar inputs occur frequently in NLG tasks. This motivates us to propose input-tuning, which fine-tunes both the continuous prompts and the input representations, leading to a more effective way to adapt unfamiliar inputs to frozen PLMs. Our proposed input-tuning is conceptually simple and empirically powerful. Experimental results on seven NLG tasks demonstrate that input-tuning is significantly and consistently better than prompt-tuning. Furthermore, on three of these tasks, input-tuning can achieve a comparable or even better performance than fine-tuning.

Yifei Li, Zeqi Lin, Shizhuo Zhang et al.

Few-shot learning is a challenging task that requires language models to generalize from limited examples. Large language models like GPT-3 and PaLM have made impressive progress in this area, but they still face difficulties in reasoning tasks such as GSM8K, a benchmark for arithmetic problems. To improve their reasoning skills, previous work has proposed to guide the language model with prompts that elicit a series of reasoning steps before giving the final answer, achieving a significant improvement on GSM8K from 17.9% to 58.1% in problem-solving rate. In this paper, we present DIVERSE (Diverse Verifier on Reasoning Step), a novel approach that further enhances the reasoning capability of language models. DIVERSE has three main components: first, it generates diverse prompts to explore different reasoning paths for the same question; second, it uses a verifier to filter out incorrect answers based on a weighted voting scheme; and third, it verifies each reasoning step individually instead of the whole chain. We evaluate DIVERSE on the latest language model code-davinci-002 and show that it achieves new state-of-the-art results on six of eight reasoning benchmarks (e.g., GSM8K 74.4% to 83.2%).

Yifei Li, Hsiao-yu Chen, Egor Larionov et al. · mit

The realism of digital avatars is crucial in enabling telepresence applications with self-expression and customization. While physical simulations can produce realistic motions for clothed humans, they require high-quality garment assets with associated physical parameters for cloth simulations. However, manually creating these assets and calibrating their parameters is labor-intensive and requires specialized expertise. Current methods focus on reconstructing geometry, but don't generate complete assets for physics-based applications. To address this gap, we propose \papername,~a novel approach that performs body and garment co-optimization using differentiable simulation. By integrating physical simulation into the optimization loop and accounting for the complex nonlinear behavior of cloth and its intricate interaction with the body, our framework recovers body and garment geometry and extracts important material parameters in a physically plausible way. Our experiments demonstrate that our approach generates realistic clothing and body shape suitable for downstream applications. We provide additional insights and results on our webpage: https://people.csail.mit.edu/liyifei/publication/diffavatar/

Tianshu Zhang, Xiang Yue, Yifei Li et al.

Semi-structured tables are ubiquitous. There has been a variety of tasks that aim to automatically interpret, augment, and query tables. Current methods often require pretraining on tables or special model architecture design, are restricted to specific table types, or have simplifying assumptions about tables and tasks. This paper makes the first step towards developing open-source large language models (LLMs) as generalists for a diversity of table-based tasks. Towards that end, we construct TableInstruct, a new dataset with a variety of realistic tables and tasks, for instruction tuning and evaluating LLMs. We further develop the first open-source generalist model for tables, TableLlama, by fine-tuning Llama 2 (7B) with LongLoRA to address the long context challenge. We experiment under both in-domain setting and out-of-domain setting. On 7 out of 8 in-domain tasks, TableLlama achieves comparable or better performance than the SOTA for each task, despite the latter often has task-specific design. On 6 out-of-domain datasets, it achieves 5-44 absolute point gains compared with the base model, showing that training on TableInstruct enhances the model's generalizability. We open-source our dataset and trained model to boost future work on developing open generalist models for tables.

Liane Makatura, Michael Foshey, Bohan Wang et al.

The advancement of Large Language Models (LLMs), including GPT-4, provides exciting new opportunities for generative design. We investigate the application of this tool across the entire design and manufacturing workflow. Specifically, we scrutinize the utility of LLMs in tasks such as: converting a text-based prompt into a design specification, transforming a design into manufacturing instructions, producing a design space and design variations, computing the performance of a design, and searching for designs predicated on performance. Through a series of examples, we highlight both the benefits and the limitations of the current LLMs. By exposing these limitations, we aspire to catalyze the continued improvement and progression of these models.

Harrison Jesse Smith, Qingyuan Zheng, Yifei Li et al.

Children's drawings have a wonderful inventiveness, creativity, and variety to them. We present a system that automatically animates children's drawings of the human figure, is robust to the variance inherent in these depictions, and is simple and straightforward enough for anyone to use. We demonstrate the value and broad appeal of our approach by building and releasing the Animated Drawings Demo, a freely available public website that has been used by millions of people around the world. We present a set of experiments exploring the amount of training data needed for fine-tuning, as well as a perceptual study demonstrating the appeal of a novel twisted perspective retargeting technique. Finally, we introduce the Amateur Drawings Dataset, a first-of-its-kind annotated dataset, collected via the public demo, containing over 178,000 amateur drawings and corresponding user-accepted character bounding boxes, segmentation masks, and joint location annotations.

Miao Xiong, Zhiyuan Hu, Xinyang Lu et al.

Empowering large language models to accurately express confidence in their answers is essential for trustworthy decision-making. Previous confidence elicitation methods, which primarily rely on white-box access to internal model information or model fine-tuning, have become less suitable for LLMs, especially closed-source commercial APIs. This leads to a growing need to explore the untapped area of black-box approaches for LLM uncertainty estimation. To better break down the problem, we define a systematic framework with three components: prompting strategies for eliciting verbalized confidence, sampling methods for generating multiple responses, and aggregation techniques for computing consistency. We then benchmark these methods on two key tasks-confidence calibration and failure prediction-across five types of datasets (e.g., commonsense and arithmetic reasoning) and five widely-used LLMs including GPT-4 and LLaMA 2 Chat. Our analysis uncovers several key insights: 1) LLMs, when verbalizing their confidence, tend to be overconfident, potentially imitating human patterns of expressing confidence. 2) As model capability scales up, both calibration and failure prediction performance improve. 3) Employing our proposed strategies, such as human-inspired prompts, consistency among multiple responses, and better aggregation strategies can help mitigate this overconfidence from various perspectives. 4) Comparisons with white-box methods indicate that while white-box methods perform better, the gap is narrow, e.g., 0.522 to 0.605 in AUROC. Despite these advancements, none of these techniques consistently outperform others, and all investigated methods struggle in challenging tasks, such as those requiring professional knowledge, indicating significant scope for improvement. We believe this study can serve as a strong baseline and provide insights for eliciting confidence in black-box LLMs.

Yifei Li, Guanyi Chen, Tingting He

Zero Pronouns (ZPs) are a pervasive linguistic phenomenon in pro-drop languages such as Chinese and have long posed a challenge for natural language processing systems. Although Large Language Models (LLMs) perform well on many Chinese language tasks, their ability to process ZPs remains poorly understood. We conduct a systematic investigation of LLMs' handling of Chinese ZPs through a sequence of linguistically motivated tasks, including identification, referentiality classification, referential type classification, resolution, and translation. A diverse set of LLMs is evaluated across all tasks. Our results show that Chinese ZPs remain highly challenging for current LLMs, particularly for upstream tasks such as identification and referentiality classification. Performance on downstream tasks, such as ZP translation, is also consistently low: even state-of-the-art reasoning-oriented LLMs correctly translate fewer than half of Chinese ZPs into English.

Peiyi Wang, Lei Li, Zhihong Shao et al. · pku

In this paper, we present an innovative process-oriented math process reward model called \textbf{Math-Shepherd}, which assigns a reward score to each step of math problem solutions. The training of Math-Shepherd is achieved using automatically constructed process-wise supervision data, breaking the bottleneck of heavy reliance on manual annotation in existing work. We explore the effectiveness of Math-Shepherd in two scenarios: 1) \textit{Verification}: Math-Shepherd is utilized for reranking multiple outputs generated by Large Language Models (LLMs); 2) \textit{Reinforcement Learning}: Math-Shepherd is employed to reinforce LLMs with step-by-step Proximal Policy Optimization (PPO). With Math-Shepherd, a series of open-source LLMs demonstrates exceptional performance. For instance, the step-by-step PPO with Math-Shepherd significantly improves the accuracy of Mistral-7B (77.9\%$\to$84.1\% on GSM8K and 28.6\%$\to$33.0\% on MATH). The accuracy can be further enhanced to 89.1\% and 43.5\% on GSM8K and MATH with the verification of Math-Shepherd, respectively. We believe that automatic process supervision holds significant potential for the future evolution of LLMs.

Team Wan, Ang Wang, Baole Ai et al.

This report presents Wan, a comprehensive and open suite of video foundation models designed to push the boundaries of video generation. Built upon the mainstream diffusion transformer paradigm, Wan achieves significant advancements in generative capabilities through a series of innovations, including our novel VAE, scalable pre-training strategies, large-scale data curation, and automated evaluation metrics. These contributions collectively enhance the model's performance and versatility. Specifically, Wan is characterized by four key features: Leading Performance: The 14B model of Wan, trained on a vast dataset comprising billions of images and videos, demonstrates the scaling laws of video generation with respect to both data and model size. It consistently outperforms the existing open-source models as well as state-of-the-art commercial solutions across multiple internal and external benchmarks, demonstrating a clear and significant performance superiority. Comprehensiveness: Wan offers two capable models, i.e., 1.3B and 14B parameters, for efficiency and effectiveness respectively. It also covers multiple downstream applications, including image-to-video, instruction-guided video editing, and personal video generation, encompassing up to eight tasks. Consumer-Grade Efficiency: The 1.3B model demonstrates exceptional resource efficiency, requiring only 8.19 GB VRAM, making it compatible with a wide range of consumer-grade GPUs. Openness: We open-source the entire series of Wan, including source code and all models, with the goal of fostering the growth of the video generation community. This openness seeks to significantly expand the creative possibilities of video production in the industry and provide academia with high-quality video foundation models. All the code and models are available at https://github.com/Wan-Video/Wan2.1.

Yuxin Zuo, Shang Qu, Yifei Li et al. · tsinghua

We introduce MedXpertQA, a highly challenging and comprehensive benchmark to evaluate expert-level medical knowledge and advanced reasoning. MedXpertQA includes 4,460 questions spanning 17 specialties and 11 body systems. It includes two subsets, Text for text evaluation and MM for multimodal evaluation. Notably, MM introduces expert-level exam questions with diverse images and rich clinical information, including patient records and examination results, setting it apart from traditional medical multimodal benchmarks with simple QA pairs generated from image captions. MedXpertQA applies rigorous filtering and augmentation to address the insufficient difficulty of existing benchmarks like MedQA, and incorporates specialty board questions to improve clinical relevance and comprehensiveness. We perform data synthesis to mitigate data leakage risk and conduct multiple rounds of expert reviews to ensure accuracy and reliability. We evaluate 18 leading models on \benchmark. Moreover, medicine is deeply connected to real-world decision-making, providing a rich and representative setting for assessing reasoning abilities beyond mathematics and code. To this end, we develop a reasoning-oriented subset to facilitate the assessment of o1-like models. Code and data are available at: https://github.com/TsinghuaC3I/MedXpertQA

Yifei Li, Weidong Guo, Lingling Zhang et al.

Long-term conversational memory is a core capability for LLM-based dialogue systems, yet existing benchmarks and evaluation protocols primarily focus on surface-level factual recall. In realistic interactions, appropriate responses often depend on implicit constraints such as user state, goals, or values that are not explicitly queried later. To evaluate this setting, we introduce \textbf{LoCoMo-Plus}, a benchmark for assessing cognitive memory under cue--trigger semantic disconnect, where models must retain and apply latent constraints across long conversational contexts. We further show that conventional string-matching metrics and explicit task-type prompting are misaligned with such scenarios, and propose a unified evaluation framework based on constraint consistency. Experiments across diverse backbone models, retrieval-based methods, and memory systems demonstrate that cognitive memory remains challenging and reveals failures not captured by existing benchmarks. Our code and evaluation framework are publicly available at: https://github.com/xjtuleeyf/Locomo-Plus.

Yifei Li, Ruizhe Fu, Huihang Liu et al.

As the demand for mass customization increases, manufacturing systems must become more flexible and adaptable to produce personalized products efficiently. Additive manufacturing (AM) enhances production adaptability by enabling on-demand fabrication of customized components directly from digital models, but its flexibility remains constrained by fixed equipment layouts. Integrating mobile robots addresses this limitation by allowing manufacturing resources to move and adapt to changing production requirements. Mobile AM Robots (MAMbots) combine AM with mobile robotics to produce and transport components within dynamic manufacturing environments. However, the dynamic manufacturing environments introduce challenges for MAMbots. Disturbances such as obstacles and uneven terrain can disrupt navigation stability, which in turn affects printing accuracy and surface quality. This work proposes a universal mobile printing-and-delivery platform that couples navigation and material deposition, addressing the limitations of earlier frameworks that treated these processes separately. A real-time control framework is developed to plan and control the robot's navigation, ensuring safe motion, obstacle avoidance, and path stability while maintaining print quality. The closed-loop integration of sensing, mobility, and manufacturing provides real-time feedback for motion and process control, enabling MAMbots to make autonomous decisions in dynamic environments. The framework is validated through simulations and real-world experiments that test its adaptability to trajectory variations and external disturbances. Coupled navigation and printing together enable MAMbots to plan safe, adaptive trajectories, improving flexibility and adaptability in manufacturing.

Chaoqi Zhen, Yanlei Shang, Xiangyu Liu et al.

Natural Language Processing (NLP) has been revolutionized by the use of Pre-trained Language Models (PLMs) such as BERT. Despite setting new records in nearly every NLP task, PLMs still face a number of challenges including poor interpretability, weak reasoning capability, and the need for a lot of expensive annotated data when applied to downstream tasks. By integrating external knowledge into PLMs, \textit{\underline{K}nowledge-\underline{E}nhanced \underline{P}re-trained \underline{L}anguage \underline{M}odels} (KEPLMs) have the potential to overcome the above-mentioned limitations. In this paper, we examine KEPLMs systematically through a series of studies. Specifically, we outline the common types and different formats of knowledge to be integrated into KEPLMs, detail the existing methods for building and evaluating KEPLMS, present the applications of KEPLMs in downstream tasks, and discuss the future research directions. Researchers will benefit from this survey by gaining a quick and comprehensive overview of the latest developments in this field.

Yifei Li, Erik-Jan van Kampen

The symmetry of dynamical systems can be exploited for state-transition prediction and to facilitate control policy optimization. This paper leverages system symmetry to develop sample-efficient offline reinforcement learning (RL) approaches. Under the symmetry assumption for a Markov Decision Process (MDP), a symmetric data augmentation method is proposed. The augmented samples are integrated into the dataset of Deep Deterministic Policy Gradient (DDPG) to enhance its coverage rate of the state-action space. Furthermore, sample utilization efficiency is improved by introducing a second critic trained on the augmented samples, resulting in a dual-critic structure. The aircraft's model is verified to be symmetric, and flight control simulations demonstrate accelerated policy convergence when augmented samples are employed.

Yifei Li, Wenzhao Zheng, Yanran Zhang et al.

The misuse of AI-driven video generation technologies has raised serious social concerns, highlighting the urgent need for reliable AI-generated video detectors. However, most existing methods are limited to binary classification and lack the necessary explanations for human interpretation. In this paper, we present Skyra, a specialized multimodal large language model (MLLM) that identifies human-perceivable visual artifacts in AI-generated videos and leverages them as grounded evidence for both detection and explanation. To support this objective, we construct ViF-CoT-4K for Supervised Fine-Tuning (SFT), which represents the first large-scale AI-generated video artifact dataset with fine-grained human annotations. We then develop a two-stage training strategy that systematically enhances our model's spatio-temporal artifact perception, explanation capability, and detection accuracy. To comprehensively evaluate Skyra, we introduce ViF-Bench, a benchmark comprising 3K high-quality samples generated by over ten state-of-the-art video generators. Extensive experiments demonstrate that Skyra surpasses existing methods across multiple benchmarks, while our evaluation yields valuable insights for advancing explainable AI-generated video detection.

Yifei Li, Haoyuan He, Yu Zheng et al.

The accessibility surge and abuse risks of user-friendly image editing models have created an urgent need for generalizable, up-to-date methods for Image Manipulation Detection and Localization (IMDL). Current IMDL research typically uses cross-dataset evaluation, where models trained on one benchmark are tested on others. However, this simplified evaluation approach conceals the fragility of existing methods when handling diverse AI-generated content, leading to misleading impressions of progress. This paper challenges this illusion by proposing NeXT-IMDL, a large-scale diagnostic benchmark designed not just to collect data, but to probe the generalization boundaries of current detectors systematically. Specifically, NeXT-IMDL categorizes AIGC-based manipulations along four fundamental axes: editing models, manipulation types, content semantics, and forgery granularity. Built upon this, NeXT-IMDL implements five rigorous cross-dimension evaluation protocols. Our extensive experiments on 11 representative models reveal a critical insight: while these models perform well in their original settings, they exhibit systemic failures and significant performance degradation when evaluated under our designed protocols that simulate real-world, various generalization scenarios. By providing this diagnostic toolkit and the new findings, we aim to advance the development towards building truly robust, next-generation IMDL models.

Minyan Luo, Yuxin Zhang, Yifei Li et al.

Narrative-driven product photography has become a prevalent paradigm in modern marketing, as coherent visual storytelling helps convey product value and establishes emotional engagement with consumers. However, existing image generation methods do not support structured narrative planning or cross-panel coordination, often resulting in weak storytelling and visual incoherence. In practice, narrative product photography is commonly presented as multi-grid collages, where multiple views or scenes jointly communicate a product narrative. To ensure visual consistency across grids and aesthetic harmony of the overall composition, we generate the collage as a single unified image rather than composing independently synthesized panels. We propose a self-reasoning agentic framework for narrative product grid collage generation. Given a product packshot and its name, the system first constructs a Product Narrative Framework that explicitly represents the product's identity, usage context, and situational environment, and translates it into complementary grids governed by a shared visual style. Constraint-aware prompts are then compiled and fed to a generation model that synthesizes the collage jointly. The generated output is evaluated on both content validity and photography quality, with explicit gates determining whether to proceed or refine. When evaluation fails, the system performs failure attribution and applies targeted refinement, enabling progressive improvement through iterative self-reflection. Experiments demonstrate that our framework consistently improves aesthetic quality, narrative richness, and visual coherence, compared to direct prompting baselines.

Yifei Li, Junbo Niu, Ziyang Miao et al. · pku, tsinghua

Temporal Awareness, the ability to reason dynamically based on the timestamp when a question is raised, is the key distinction between offline and online video LLMs. Unlike offline models, which rely on complete videos for static, post hoc analysis, online models process video streams incrementally and dynamically adapt their responses based on the timestamp at which the question is posed. Despite its significance, temporal awareness has not been adequately evaluated in existing benchmarks. To fill this gap, we present OVO-Bench (Online-VideO-Benchmark), a novel video benchmark that emphasizes the importance of timestamps for advanced online video understanding capability benchmarking. OVO-Bench evaluates the ability of video LLMs to reason and respond to events occurring at specific timestamps under three distinct scenarios: (1) Backward tracing: trace back to past events to answer the question. (2) Real-time understanding: understand and respond to events as they unfold at the current timestamp. (3) Forward active responding: delay the response until sufficient future information becomes available to answer the question accurately. OVO-Bench comprises 12 tasks, featuring 644 unique videos and approximately human-curated 2,800 fine-grained meta-annotations with precise timestamps. We combine automated generation pipelines with human curation. With these high-quality samples, we further developed an evaluation pipeline to systematically query video LLMs along the video timeline. Evaluations of nine Video-LLMs reveal that, despite advancements on traditional benchmarks, current models struggle with online video understanding, showing a significant gap compared to human agents. We hope OVO-Bench will drive progress in video LLMs and inspire future research in online video reasoning. Our benchmark and code can be accessed at https://github.com/JoeLeelyf/OVO-Bench.

Muye Huang, Lingling Zhang, Yifei Li et al.

Charts are high-density visual carriers of complex data and medium for information extraction and analysis. Due to the need for precise and complex visual reasoning, automated chart understanding poses a significant challenge to existing Multimodal Large Language Models (MLLMs). Many MLLMs trained with reinforcement learning (RL) face the challenge of credit assignment. Their advantage estimation, typically performed at the trajectory level, cannot distinguish between correct and incorrect reasoning steps within a single generated response. To address this limitation, we introduce SketchVL, a novel MLLM that optimized with FinePO, a new RL algorithm designed for fine-grained credit assignment within each trajectory. SketchVL's methodology involves drawing its intermediate reasoning steps as markers on the image and feeding the annotated image back to itself, creating a robust, multi-step reasoning process. During training, the FinePO algorithm leverages a Fine-grained Process Reward Model (FinePRM) to score each drawing action within a trajectory, thereby precisely assigning credit for each step. This mechanism allows FinePO to more strongly reward correct tokens when a trajectory is globally successful, and more heavily penalize incorrect tokens when the trajectory is globally suboptimal, thus achieving fine-grained reinforcement signals. Experiments show that SketchVL learns to align its step-level behavior with the FinePRM, achieving an average performance gain of 7.23\% over its base model across chart datasets, natural image datasets, and mathematics, providing a promising new direction for training powerful reasoning models.

Hanane Nour Moussa, Yifei Li, Zhuoyang Li et al.

Despite recent progress in language models and agents for scientific data-driven discovery, further advancing their capabilities is held back by the absence of verifiable environments representing real-world scientific tasks.To fill this gap, we introduce D3-Gym, the first automatically constructed dataset with verifiable environments for scientific Data-Driven Discovery. D3-Gym comprises (1) 565 tasks sourced from 239 real scientific repositories across four disciplines where (2) each task is equipped with a natural language instruction, an executable environment with pre-installed dependencies, input dataset and artifact previews, a reference code solution, and an automatically synthesized evaluation script. Rigorous evaluation of the quality of the verification signal in D3-Gym confirms that our evaluation scripts achieve 87.5% agreement with human-annotated gold standards and strong alignment in domain-specific evaluation logic, showing their scientific soundness. Further, training on trajectories sampled from D3-Gym yields consistent and substantial gains across Qwen3 models of varying sizes on ScienceAgentBench, boosting Qwen3-32B by 7.8 absolute points and substantially shrinking the gap with strong proprietary models. All D3-Gym artifacts (environments, creation workflow, trajectories, and models) can be found at https://github.com/OSU-NLP-Group/D3-Gym.

Yuxin Li, Xiangyu Zhang, Yifei Li et al.

Speech is a scalable and non-invasive biomarker for early mental health screening. However, widely used depression datasets like DAIC-WOZ exhibit strong coupling between linguistic sentiment and diagnostic labels, encouraging models to learn semantic shortcuts. As a result, model robustness may be compromised in real-world scenarios, such as Camouflaged Depression, where individuals maintain socially positive or neutral language despite underlying depressive states. To mitigate this semantic bias, we propose DepFlow, a three-stage depression-conditioned text-to-speech framework. First, a Depression Acoustic Encoder learns speaker- and content-invariant depression embeddings through adversarial training, achieving effective disentanglement while preserving depression discriminability (ROC-AUC: 0.693). Second, a flow-matching TTS model with FiLM modulation injects these embeddings into synthesis, enabling control over depressive severity while preserving content and speaker identity. Third, a prototype-based severity mapping mechanism provides smooth and interpretable manipulation across the depression continuum. Using DepFlow, we construct a Camouflage Depression-oriented Augmentation (CDoA) dataset that pairs depressed acoustic patterns with positive/neutral content from a sentiment-stratified text bank, creating acoustic-semantic mismatches underrepresented in natural data. Evaluated across three depression detection architectures, CDoA improves macro-F1 by 9%, 12%, and 5%, respectively, consistently outperforming conventional augmentation strategies in depression Detection. Beyond enhancing robustness, DepFlow provides a controllable synthesis platform for conversational systems and simulation-based evaluation, where real clinical data remains limited by ethical and coverage constraints.

Yanran Zhang, Wenzhao Zheng, Yifei Li et al.

In recent years, significant progress has been made in both image generation and generated image detection. Despite their rapid, yet largely independent, development, these two fields have evolved distinct architectural paradigms: the former predominantly relies on generative networks, while the latter favors discriminative frameworks. A recent trend in both domains is the use of adversarial information to enhance performance, revealing potential for synergy. However, the significant architectural divergence between them presents considerable challenges. Departing from previous approaches, we propose UniGenDet: a Unified generative-discriminative framework for co-evolutionary image Generation and generated image Detection. To bridge the task gap, we design a symbiotic multimodal self-attention mechanism and a unified fine-tuning algorithm. This synergy allows the generation task to improve the interpretability of authenticity identification, while authenticity criteria guide the creation of higher-fidelity images. Furthermore, we introduce a detector-informed generative alignment mechanism to facilitate seamless information exchange. Extensive experiments on multiple datasets demonstrate that our method achieves state-of-the-art performance. Code: \href{https://github.com/Zhangyr2022/UniGenDet}{https://github.com/Zhangyr2022/UniGenDet}.

Chen Liang, Donghua Yang, Yutong Zhao et al.

Structural missingness breaks 'just impute and train': values can be undefined by causal or logical constraints, and the mask may depend on observed variables, unobserved variables (MNAR), and other missingness indicators. It simultaneously brings (i) a catch-22 situation with causal loop, prediction needs the missing features, yet inferring them depends on the missingness mechanism, (ii) under MNAR, the unseen are different, the missing part can come from a shifted distribution, and (iii) plug-in imputation, a single fill-in can lock in uncertainty and yield overconfident, biased decisions. In the Bayesian view, prediction via the posterior predictive distribution integrates over the full model posterior uncertainty, rather than relying on a single point estimate. This framework decouples (i) learning an in-model missing-value posterior from (ii) label prediction by optimizing the predictive posterior distribution, enabling posterior integration. This decoupling yields an in-model almost-free-lunch: once the posterior is learned, prediction is plug-and-play while preserving uncertainty propagation. It achieves SOTA on 43 classification and 15 imputation benchmarks, with finite-sample near Bayes-optimality guarantees under our SCM prior.

Jun Liu, Xiaohui Zhong, Kai Zheng et al.

Tropical cyclone (TC) intensity forecasting remains challenging as current numerical and AI-based weather models fail to satisfactorily represent extreme TC structure and intensity. Although intensity time-series forecasting has achieved significant advances, it outputs intensity sequences rather than the three-dimensional inner-core fine-scale structure and physical mechanisms governing TC evolution. High-resolution numerical simulations can capture these features but remain computationally expensive and inefficient for large-scale operational applications. Here we present 3DTCR, a physics-based generative framework combining physical constraints with generative AI efficiency for 3D TC structure reconstruction. Trained on a six-year, 3-km-resolution moving-domain WRF dataset, 3DTCR enables region-adaptive vortex-following reconstruction using conditional Flow Matching(CFM), optimized via latent domain adaptation and two-stage transfer learning. The framework mitigates limitations imposed by low-resolution targets and over-smoothed forecasts, improving the representation of TC inner-core structure and intensity while maintaining track stability. Results demonstrate that 3DTCR outperforms the ECMWF high-resolution forecasting system (ECMWF-HRES) in TC intensity prediction at nearly all lead times up to 5 days and reduces the RMSE of maximum WS10M by 36.5% relative to its FuXi inputs. These findings highlight 3DTCR as a physics-based generative framework that efficiently resolves fine-scale structures at lower computational cost, which may offer a promising avenue for improving TC intensity forecasting.

Yifei Li, Richong Zhang, Wanyu Tu et al.

Legal judgments may contain errors due to the complexity of case circumstances and the abstract nature of legal concepts, while existing appellate review mechanisms face efficiency pressures from a surge in case volumes. Although current legal AI research focuses on tasks like judgment prediction and legal document generation, the task of judgment review differs fundamentally in its objectives and paradigm: it centers on detecting, classifying, and correcting errors after a judgment is issued, constituting anomaly detection rather than prediction or generation. To address this research gap, we introduce a novel task APPELLATE REVIEW, aiming to assess models' diagnostic reasoning and reliability in legal practice. We also construct a novel dataset benchmark AR-BENCH, which comprises 8,700 finely annotated decisions and 34,617 supplementary corpora. By evaluating 14 large language models, we reveal critical limitations in existing models' ability to identify legal application errors, providing empirical evidence for future improvements.

Zehao Chen, Tianxiang Ai, Yifei Li et al.

Ensemble learning of LLMs has emerged as a promising alternative to enhance performance, but existing approaches typically treat models as black boxes, combining the inputs or final outputs while overlooking the rich internal representations and interactions across models.In this work, we introduce LLMBoost, a novel ensemble fine-tuning framework that breaks this barrier by explicitly leveraging intermediate states of LLMs. Inspired by the boosting paradigm, LLMBoost incorporates three key innovations. First, a cross-model attention mechanism enables successor models to access and fuse hidden states from predecessors, facilitating hierarchical error correction and knowledge transfer. Second, a chain training paradigm progressively fine-tunes connected models with an error-suppression objective, ensuring that each model rectifies the mispredictions of its predecessor with minimal additional computation. Third, a near-parallel inference paradigm design pipelines hidden states across models layer by layer, achieving inference efficiency approaching single-model decoding. We further establish the theoretical foundations of LLMBoost, proving that sequential integration guarantees monotonic improvements under bounded correction assumptions. Extensive experiments on commonsense reasoning and arithmetic reasoning tasks demonstrate that LLMBoost consistently boosts accuracy while reducing inference latency.

Weizhu Bao, Yifei Li, Wenjun Ying et al.

We propose a structure-preserving parametric approximation for geometric flows with general anisotropic effects. By introducing a hyperparameter $α$, we construct a unified surface energy matrix $\hat{\boldsymbol{G}}_k^α(θ)$ that encompasses all existing formulations of surface energy matrices, and apply it to anisotropic curvature flow. We prove that $α=-1$ is the unique choice achieving optimal energy stability under the necessary and sufficient condition $3\hatγ(θ)\geq\hatγ(θ-π)$, while all other $α\neq-1$ require strictly stronger conditions. The framework extends naturally to general anisotropic geometric flows through a unified velocity discretization that ensures energy stability. Numerical experiments validate the theoretical optimality of $α=-1$ and demonstrate the effectiveness and robustness.

Rongman Xu, Yifei Li, Tianzhe Zhao et al.

Large Language Models (LLMs) have demonstrated remarkable abilities in reasoning. However, maximizing their potential through inference-time scaling faces challenges in trade-off between sampling budget and reasoning quality. Current strategies remain inefficient as they typically treat sampling width and depth as orthogonal objectives, where width consensus methods risk reinforcing hallucinations, while depth pruning mechanisms prematurely truncate complex yet valid reasoning chains. Therefore, we propose Dual-Dimensional Consistency (DDC), a unified framework that bridges path quality with adaptive termination. By coupling Confidence-Weighted Bayesian protocol with a Trend-Aware Stratified Pruning, our method ensures that computational resources are concentrated on high quality reasoning paths, filtering hallucinations while accelerating consensus. Evaluations across five benchmarks demonstrate that this approach reduces token consumption by over 10 times while maintaining or exceeding the accuracy of strong baselines across various LLMs.

Kyungmi Na, Yifei Li, Xinyi Yang et al.

In biological systems, sensing is not performed by the brain alone: the body deforms, vibrates, and filters external stimuli before they are transduced into neural signals. In engineered systems, this processing burden is placed largely on electronics and computation, while the mechanical body is usually designed only for strength and stability. Here, we present sensing intelligence as a trainable property of the body. We show that the geometry of a metamaterial can be optimized to reshape external stimuli into internal signals that are easier for a neural network to interpret. Rather than hand-designing this physical preprocessing, we let the neural network train its own body for sensing by backpropagating the sensing loss to the body's design parameters through differentiable simulation. Across numerical and experimental sensing scenarios, the optimized body improves sensing accuracy by up to fivefold or reduces the number of required electronic sensors by nearly an order of magnitude.

Zehao Chen, Gongxun Li, Tianxiang Ai et al.

As post-training optimization becomes central to improving large language models, we observe a persistent saturation bottleneck: once models grow highly confident, further training yields diminishing returns. While existing methods continue to reinforce target predictions, we find that informative supervision signals remain latent in models' own historical weak states. Motivated by this observation, we propose WMSS (Weak Agents Can Make Strong Agents Stronger), a post-training paradigm that leverages weak checkpoints to guide continued optimization. By identifying recoverable learning gaps via entropy dynamics and reinforcing them through compensatory learning, WMSS enables strong agents to improve beyond conventional post-training saturation. Experiments on mathematical reasoning and code generation datasets show that agents trained with our approach achieve effective performance improvements, while incurring zero additional inference cost.

Yifei Li, Lingling Zhang, Hang Yan et al.

Traditional knowledge graph (KG) embedding methods aim to represent entities and relations in a low-dimensional space, primarily focusing on static graphs. However, real-world KGs are dynamically evolving with the constant addition of entities, relations and facts. To address such dynamic nature of KGs, several continual knowledge graph embedding (CKGE) methods have been developed to efficiently update KG embeddings to accommodate new facts while maintaining learned knowledge. As KGs grow at different rates and scales in real-world scenarios, existing CKGE methods often fail to consider the varying scales of updates and lack systematic evaluation throughout the entire update process. In this paper, we propose SAGE, a scale-aware gradual evolution framework for CKGE. Specifically, SAGE firstly determine the embedding dimensions based on the update scales and expand the embedding space accordingly. The Dynamic Distillation mechanism is further employed to balance the preservation of learned knowledge and the incorporation of new facts. We conduct extensive experiments on seven benchmarks, and the results show that SAGE consistently outperforms existing baselines, with a notable improvement of 1.38% in MRR, 1.25% in H@1 and 1.6% in H@10. Furthermore, experiments comparing SAGE with methods using fixed embedding dimensions show that SAGE achieves optimal performance on every snapshot, demonstrating the importance of adaptive embedding dimensions in CKGE. The codes of SAGE are publicly available at: https://github.com/lyfxjtu/Dynamic-Embedding.

Kaiwen Shi, Yifei Li, Binh Ho et al.

In recent years, machine learning algorithms have achieved much success in segmenting lesions across various tissues. There is, however, not one satisfying model that works well on all tissue types universally. In response to this need, we attempt to train a model that 1) works well on all tissue types, and 2) is capable of still performing fast inferences. To this end, we design our architectures, test multiple existing architectures, compare their results, and settle upon SwinUnet. We document our rationales, successes, and failures. Finally, we propose some further directions that we think are worth exploring. codes: https://github.com/KWFredShi/ULS2023NGKD.git

Pan Zhang, Xiaoyi Dong, Yuhang Cao et al. · pku, tsinghua

Creating AI systems that can interact with environments over long periods, similar to human cognition, has been a longstanding research goal. Recent advancements in multimodal large language models (MLLMs) have made significant strides in open-world understanding. However, the challenge of continuous and simultaneous streaming perception, memory, and reasoning remains largely unexplored. Current MLLMs are constrained by their sequence-to-sequence architecture, which limits their ability to process inputs and generate responses simultaneously, akin to being unable to think while perceiving. Furthermore, relying on long contexts to store historical data is impractical for long-term interactions, as retaining all information becomes costly and inefficient. Therefore, rather than relying on a single foundation model to perform all functions, this project draws inspiration from the concept of the Specialized Generalist AI and introduces disentangled streaming perception, reasoning, and memory mechanisms, enabling real-time interaction with streaming video and audio input. The proposed framework InternLM-XComposer2.5-OmniLive (IXC2.5-OL) consists of three key modules: (1) Streaming Perception Module: Processes multimodal information in real-time, storing key details in memory and triggering reasoning in response to user queries. (2) Multi-modal Long Memory Module: Integrates short-term and long-term memory, compressing short-term memories into long-term ones for efficient retrieval and improved accuracy. (3) Reasoning Module: Responds to queries and executes reasoning tasks, coordinating with the perception and memory modules. This project simulates human-like cognition, enabling multimodal large language models to provide continuous and adaptive service over time.

Yifei Li, Xiang Yue, Zeyi Liao et al.

Modern generative search engines enhance the reliability of large language model (LLM) responses by providing cited evidence. However, evaluating the answer's attribution, i.e., whether every claim within the generated responses is fully supported by its cited evidence, remains an open problem. This verification, traditionally dependent on costly human evaluation, underscores the urgent need for automatic attribution evaluation methods. To bridge the gap in the absence of standardized benchmarks for these methods, we present AttributionBench, a comprehensive benchmark compiled from various existing attribution datasets. Our extensive experiments on AttributionBench reveal the challenges of automatic attribution evaluation, even for state-of-the-art LLMs. Specifically, our findings show that even a fine-tuned GPT-3.5 only achieves around 80% macro-F1 under a binary classification formulation. A detailed analysis of more than 300 error cases indicates that a majority of failures stem from the model's inability to process nuanced information, and the discrepancy between the information the model has access to and that human annotators do.

Jintao Xu, Yifei Li, Wenxun Xing

We propose both serial and parallel proximal (linearized) alternating direction method of multipliers (ADMM) algorithms for training residual neural networks. In contrast to backpropagation-based approaches, our methods inherently mitigate the exploding gradient issue and are well-suited for parallel and distributed training through regional updates. Theoretically, we prove that the proposed algorithms converge at an R-linear (sublinear) rate for both the iteration points and the objective function values. These results hold without imposing stringent constraints on network width, depth, or training data size. Furthermore, we theoretically analyze our parallel/distributed ADMM algorithms, highlighting their reduced time complexity and lower per-node memory consumption. To facilitate practical deployment, we develop a control protocol for parallel ADMM implementation using Python's multiprocessing and interprocess communication. Experimental results validate the proposed ADMM algorithms, demonstrating rapid and stable convergence, improved performance, and high computational efficiency. Finally, we highlight the improved scalability and efficiency achieved by our parallel ADMM training strategy.

Junhao Cheng, Baiqiao Yin, Kaixin Cai et al.

Recent advances in diffusion models can generate high-quality and stunning images from text. However, multi-turn image generation, which is of high demand in real-world scenarios, still faces challenges in maintaining semantic consistency between images and texts, as well as contextual consistency of the same subject across multiple interactive turns. To address this issue, we introduce TheaterGen, a training-free framework that integrates large language models (LLMs) and text-to-image (T2I) models to provide the capability of multi-turn image generation. Within this framework, LLMs, acting as a "Screenwriter", engage in multi-turn interaction, generating and managing a standardized prompt book that encompasses prompts and layout designs for each character in the target image. Based on these, Theatergen generate a list of character images and extract guidance information, akin to the "Rehearsal". Subsequently, through incorporating the prompt book and guidance information into the reverse denoising process of T2I diffusion models, Theatergen generate the final image, as conducting the "Final Performance". With the effective management of prompt books and character images, TheaterGen significantly improves semantic and contextual consistency in synthesized images. Furthermore, we introduce a dedicated benchmark, CMIGBench (Consistent Multi-turn Image Generation Benchmark) with 8000 multi-turn instructions. Different from previous multi-turn benchmarks, CMIGBench does not define characters in advance. Both the tasks of story generation and multi-turn editing are included on CMIGBench for comprehensive evaluation. Extensive experimental results show that TheaterGen outperforms state-of-the-art methods significantly. It raises the performance bar of the cutting-edge Mini DALLE 3 model by 21% in average character-character similarity and 19% in average text-image similarity.

Yifei Li, Haixu Wu, Zeyi Xu et al.

Learning-based methods have made significant progress in physics simulation, typically approximating dynamics with a monolithic end-to-end optimized neural network. Although these models offer an effective way to simulation, they may lose essential features compared to traditional numerical simulators, such as physical interpretability and reliability. Drawing inspiration from classical simulators that operate in a modular fashion, this paper presents Neural Modular Physics (NMP) for elastic simulation, which combines the approximation capacity of neural networks with the physical reliability of traditional simulators. Beyond the previous monolithic learning paradigm, NMP enables direct supervision of intermediate quantities and physical constraints by decomposing elastic dynamics into physically meaningful neural modules connected through intermediate physical quantities. With a specialized architecture and training strategy, our method transforms the numerical computation flow into a modular neural simulator, achieving improved physical consistency and generalizability. Experimentally, NMP demonstrates superior generalization to unseen initial conditions and resolutions, stable long-horizon simulation, better preservation of physical properties compared to other neural simulators, and greater feasibility in scenarios with unknown underlying dynamics than traditional simulators.

Boyu Gou, Zanming Huang, Yuting Ning et al. · microsoft-research

Agentic search such as Deep Research systems-where agents autonomously browse the web, synthesize information, and return comprehensive citation-backed answers-represents a major shift in how users interact with web-scale information. While promising greater efficiency and cognitive offloading, the growing complexity and open-endedness of agentic search have outpaced existing evaluation benchmarks and methodologies, which largely assume short search horizons and static answers. In this paper, we introduce Mind2Web 2, a benchmark of 130 realistic, high-quality, and long-horizon tasks that require real-time web browsing and extensive information synthesis, constructed with over 1000 hours of human labor. To address the challenge of evaluating time-varying and complex answers, we propose a novel Agent-as-a-Judge framework. Our method constructs task-specific judge agents based on a tree-structured rubric design to automatically assess both answer correctness and source attribution. We conduct a comprehensive evaluation of ten frontier agentic search systems and human performance, along with a detailed error analysis to draw insights for future development. The best-performing system, OpenAI Deep Research, can already achieve 50-70% of human performance while spending half the time, highlighting its great potential. Altogether, Mind2Web 2 provides a rigorous foundation for developing and benchmarking the next generation of agentic search systems.

Ning Ding, Shang Qu, Linhai Xie et al. · tsinghua

With the development of artificial intelligence, its contribution to science is evolving from simulating a complex problem to automating entire research processes and producing novel discoveries. Achieving this advancement requires both specialized general models grounded in real-world scientific data and iterative, exploratory frameworks that mirror human scientific methodologies. In this paper, we present PROTEUS, a fully automated system for scientific discovery from raw proteomics data. PROTEUS uses large language models (LLMs) to perform hierarchical planning, execute specialized bioinformatics tools, and iteratively refine analysis workflows to generate high-quality scientific hypotheses. The system takes proteomics datasets as input and produces a comprehensive set of research objectives, analysis results, and novel biological hypotheses without human intervention. We evaluated PROTEUS on 12 proteomics datasets collected from various biological samples (e.g. immune cells, tumors) and different sample types (single-cell and bulk), generating 191 scientific hypotheses. These were assessed using both automatic LLM-based scoring on 5 metrics and detailed reviews from human experts. Results demonstrate that PROTEUS consistently produces reliable, logically coherent results that align well with existing literature while also proposing novel, evaluable hypotheses. The system's flexible architecture facilitates seamless integration of diverse analysis tools and adaptation to different proteomics data types. By automating complex proteomics analysis workflows and hypothesis generation, PROTEUS has the potential to considerably accelerate the pace of scientific discovery in proteomics research, enabling researchers to efficiently explore large-scale datasets and uncover biological insights.

Yifei Li, Kang Wu, Wenming Wu et al.

We introduce a novel self-supervised learning framework that automatically learns representations from input computer-aided design (CAD) models for downstream tasks, including part classification, modeling segmentation, and machining feature recognition. To train our network, we construct a large-scale, unlabeled dataset of boundary representation (BRep) models. The success of our algorithm relies on two keycomponents. The first is a masked graph autoencoder that reconstructs randomly masked geometries and attributes of BReps for representation learning to enhance the generalization. The second is a hierarchical graph Transformer architecture that elegantly fuses global and local learning by a cross-scale mutual attention block to model long-range geometric dependencies and a graph neural network block to aggregate local topological information. After training the autoencoder, we replace its decoder with a task-specific network trained on a small amount of labeled data for downstream tasks. We conduct experiments on various tasks and achieve high performance, even with a small amount of labeled data, demonstrating the practicality and generalizability of our model. Compared to other methods, our model performs significantly better on downstream tasks with the same amount of training data, particularly when the training data is very limited.

Hang Yan, Fangzhi Xu, Rongman Xu et al. · mit

Large Language Models (LLMs) have achieved impressive performance on reasoning-intensive tasks, yet optimizing their reasoning efficiency remains an open challenge. While Test-Time Scaling (TTS) improves reasoning quality, it often leads to overthinking, wasting tokens on redundant computations. This work investigates how to efficiently and adaptively guide LLM test-time scaling without additional training. Inspired by the concept of momentum in physics, we propose Momentum Uncertainty-guided Reasoning (MUR), which dynamically allocates thinking budgets to critical reasoning steps by tracking and aggregating stepwise uncertainty over time. To support flexible inference-time control, we introduce gamma-control, a simple mechanism that tunes the reasoning budget via a single hyperparameter. We provide in-depth theoretical proof to support the superiority of MUR in terms of stability and biases. MUR is comprehensively evaluated against various TTS methods across four challenging benchmarks (MATH-500, AIME24, AIME25, and GPQA-diamond) using different sizes of recent Qwen3 models (1.7B, 4B, and 8B). Results demonstrate that MUR reduces computation by over 50% on average while improving accuracy by 0.62-3.37%.

Yifei Li, Yuchen Sun, Pingchuan Ma et al. · mit

We present a novel framework to explore neural control and design of complex fluidic systems with dynamic solid boundaries. Our system features a fast differentiable Navier-Stokes solver with solid-fluid interface handling, a low-dimensional differentiable parametric geometry representation, a control-shape co-design algorithm, and gym-like simulation environments to facilitate various fluidic control design applications. Additionally, we present a benchmark of design, control, and learning tasks on high-fidelity, high-resolution dynamic fluid environments that pose challenges for existing differentiable fluid simulators. These tasks include designing the control of artificial hearts, identifying robotic end-effector shapes, and controlling a fluid gate. By seamlessly incorporating our differentiable fluid simulator into a learning framework, we demonstrate successful design, control, and learning results that surpass gradient-free solutions in these benchmark tasks.

Yifei Li, Hanane Nour Moussa, Ziru Chen et al.

Despite long-standing efforts in accelerating scientific discovery with AI, building AI co-scientists remains challenging due to limited high-quality data for training and evaluation. To tackle this data scarcity issue, we present AutoSDT, an automatic pipeline that collects high-quality coding tasks in real-world data-driven discovery workflows. AutoSDT leverages the coding capabilities and parametric knowledge of LLMs to search for diverse sources, select ecologically valid tasks, and synthesize accurate task instructions and code solutions. Using our pipeline, we construct AutoSDT-5K, a dataset of 5,404 coding tasks for data-driven discovery that covers four scientific disciplines and 756 unique Python packages. To the best of our knowledge, AutoSDT-5K is the only automatically collected and the largest open dataset for data-driven scientific discovery. Expert feedback on a subset of 256 tasks shows the effectiveness of AutoSDT: 93% of the collected tasks are ecologically valid, and 92.2% of the synthesized programs are functionally correct. Trained on AutoSDT-5K, the Qwen2.5-Coder-Instruct LLM series, dubbed AutoSDT-Coder, show substantial improvement on two challenging data-driven discovery benchmarks, ScienceAgentBench and DiscoveryBench. Most notably, AutoSDT-Coder-32B reaches the same level of performance as GPT-4o on ScienceAgentBench with a success rate of 7.8%, doubling the performance of its base model. On DiscoveryBench, it lifts the hypothesis matching score to 8.1, bringing a 17.4% relative improvement and closing the gap between open-weight models and GPT-4o.

Xiao Long, Liansheng Zhuang, Chen Shen et al.

Recently, large language models (LLMs) have demonstrated impressive performance in Knowledge Graph Question Answering (KGQA) tasks, which aim to find answers based on knowledge graphs (KGs) for natural language questions. Existing LLMs-based KGQA methods typically follow the Graph Retrieval-Augmented Generation (GraphRAG) paradigm, which first retrieves reasoning paths from the large KGs, and then generates the answers based on them. However, these methods emphasize the exploration of new optimal reasoning paths in KGs while ignoring the exploitation of historical reasoning paths, which may lead to sub-optimal reasoning paths. Additionally, the complex semantics contained in questions may lead to the retrieval of inaccurate reasoning paths. To address these issues, this paper proposes a novel and training-free framework for KGQA tasks called Reward-guided Tree Search on Graph (RTSoG). RTSoG decomposes an original question into a series of simpler and well-defined sub-questions to handle the complex semantics. Then, a Self-Critic Monte Carlo Tree Search (SC-MCTS) guided by a reward model is introduced to iteratively retrieve weighted reasoning paths as contextual knowledge. Finally, it stacks the weighted reasoning paths according to their weights to generate the final answers. Extensive experiments on four datasets demonstrate the effectiveness of RTSoG. Notably, it achieves 8.7\% and 7.0\% performance improvement over the state-of-the-art method on the GrailQA and the WebQSP respectively.

Junrui Wen, Yifei Li, Bart Selman et al.

Neural solvers have shown significant potential in solving the Traveling Salesman Problem (TSP), yet current approaches face significant challenges. Supervised learning (SL)-based solvers require large amounts of high-quality labeled data, while reinforcement learning (RL)-based solvers, though less dependent on such data, often suffer from inefficiencies. To address these limitations, we propose LocalEscaper, a novel weakly-supervised learning framework for large-scale TSP. LocalEscaper effectively combines the advantages of both SL and RL, enabling effective training on datasets with low-quality labels. To further enhance solution quality, we introduce a regional reconstruction strategy, which is the key technique of this paper and mitigates the local-optima problem common in existing local reconstruction methods. Experimental results on both synthetic and real-world datasets demonstrate that LocalEscaper outperforms existing neural solvers, achieving remarkable results.

Muye Huang, Lingling Zhang, Jie Ma et al.

Charts are high-density visualization carriers for complex data, serving as a crucial medium for information extraction and analysis. Automated chart understanding poses significant challenges to existing multimodal large language models (MLLMs) due to the need for precise and complex visual reasoning. Current step-by-step reasoning models primarily focus on text-based logical reasoning for chart understanding. However, they struggle to refine or correct their reasoning when errors stem from flawed visual understanding, as they lack the ability to leverage multimodal interaction for deeper comprehension. Inspired by human cognitive behavior, we propose ChartSketcher, a multimodal feedback-driven step-by-step reasoning method designed to address these limitations. ChartSketcher is a chart understanding model that employs Sketch-CoT, enabling MLLMs to annotate intermediate reasoning steps directly onto charts using a programmatic sketching library, iteratively feeding these visual annotations back into the reasoning process. This mechanism enables the model to visually ground its reasoning and refine its understanding over multiple steps. We employ a two-stage training strategy: a cold start phase to learn sketch-based reasoning patterns, followed by off-policy reinforcement learning to enhance reflection and generalization. Experiments demonstrate that ChartSketcher achieves promising performance on chart understanding benchmarks and general vision tasks, providing an interactive and interpretable approach to chart comprehension.