Yanyu Chen

Gengyuan Zhang, Jinhe Bi, Jindong Gu et al. · deepmind, oxford

Understanding videos is an important research topic for multimodal learning. Leveraging large-scale datasets of web-crawled video-text pairs as weak supervision has become a pre-training paradigm for learning joint representations and showcased remarkable potential in video understanding tasks. However, videos can be multi-event and multi-grained, while these video-text pairs usually contain only broad-level video captions. This raises a question: with such weak supervision, can video representation in video-language models gain the ability to distinguish even factual discrepancies in textual description and understand fine-grained events? To address this, we introduce SPOT Prober, to benchmark existing video-language models's capacities of distinguishing event-level discrepancies as an indicator of models' event understanding ability. Our approach involves extracting events as tuples (<Subject, Predicate, Object, Attribute, Timestamps>) from videos and generating false event tuples by manipulating tuple components systematically. We reevaluate the existing video-language models with these positive and negative captions and find they fail to distinguish most of the manipulated events. Based on our findings, we propose to plug in these manipulated event captions as hard negative samples and find them effective in enhancing models for event understanding.

Zheng Wu, Chengcheng Han, Zhengxi Lu et al.

Despite the rapid progress of multimodal large language models in building Graphical User Interface (GUI) agents, their real-world task completion is fundamentally bottlenecked by a lack of world knowledge about GUI operations. Existing solutions typically rely on expensive multi-agent scaffolding or conventional post-training paradigms, such as Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL). However, post-training only allows agents to implicitly absorb world knowledge through action annotations or reward signals, leading to inefficient trajectory memorization rather than genuine comprehension. Therefore, an approach that enables explicit learning of this knowledge is imperative. To this end, we propose GUI-CIDER, a mid-training method that explicitly internalizes GUI world knowledge through Causal Internalization and Density-aware Exemplar Reselection. GUI-CIDER operates in three stages: (1) data synthesis, which distills static planning and dynamic causal knowledge from GUI trajectories into text; (2) exemplar reselection, which filters the corpus by rewarding causal structures and penalizing semantic redundancy; and (3) mid-training, where the refined data is used to embed the acquired knowledge. Extensive experiments on two GUI knowledge benchmarks and three task completion benchmarks demonstrate that GUI-CIDER consistently improves both the agent's understanding of GUI operations and its task success rates.The codes are available at https://github.com/Wuzheng02/GUI-CIDER.

Meituan LongCat Team, Anchun Gui, Bei Li et al.

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

Yuwen Deng, Donghai Guan, Yanyu Chen et al.

This paper studies a practically meaningful ship detection problem from synthetic aperture radar (SAR) images by the neural network. We broadly extract different types of SAR image features and raise the intriguing question that whether these extracted features are beneficial to (1) suppress data variations (e.g., complex land-sea backgrounds, scattered noise) of real-world SAR images, and (2) enhance the features of ships that are small objects and have different aspect (length-width) ratios, therefore resulting in the improvement of ship detection. To answer this question, we propose a SAR-ship detection neural network (call SAR-ShipNet for short), by newly developing Bidirectional Coordinate Attention (BCA) and Multi-resolution Feature Fusion (MRF) based on CenterNet. Moreover, considering the varying length-width ratio of arbitrary ships, we adopt elliptical Gaussian probability distribution in CenterNet to improve the performance of base detector models. Experimental results on the public SAR-Ship dataset show that our SAR-ShipNet achieves competitive advantages in both speed and accuracy.

Dugang Liu, Pengxiang Cheng, Hong Zhu et al.

Tabular data is one of the most common data storage formats behind many real-world web applications such as retail, banking, and e-commerce. The success of these web applications largely depends on the ability of the employed machine learning model to accurately distinguish influential features from all the predetermined features in tabular data. Intuitively, in practical business scenarios, different instances should correspond to different sets of influential features, and the set of influential features of the same instance may vary in different scenarios. However, most existing methods focus on global feature selection assuming that all instances have the same set of influential features, and few methods considering instance-wise feature selection ignore the variability of influential features in different scenarios. In this paper, we first introduce a new perspective based on the influence function for instance-wise feature selection, and give some corresponding theoretical insights, the core of which is to use the influence function as an indicator to measure the importance of an instance-wise feature. We then propose a new solution for discovering instance-wise influential features in tabular data (DIWIFT), where a self-attention network is used as a feature selection model and the value of the corresponding influence function is used as an optimization objective to guide the model. Benefiting from the advantage of the influence function, i.e., its computation does not depend on a specific architecture and can also take into account the data distribution in different scenarios, our DIWIFT has better flexibility and robustness. Finally, we conduct extensive experiments on both synthetic and real-world datasets to validate the effectiveness of our DIWIFT.

Yanyu Chen, Jiyue Jiang, Dianzhi Yu et al.

The evolution of Large Language Model (LLM) reasoning is bottlenecked by the scarcity of high-quality process data. While self-alignment via endogenous rewards offers a solution, mining valid supervision faces three challenges: (1) Label Noise via Mimetic Bias, where rewards prioritize statistical likelihood over logical truth, creating a "correctness illusion" that masks compounding errors; (2) Coarse-Grained Supervision, where sparse global outcomes (e.g., in GRPO) fail to provide granular guidance, treating reasoning chains as monolithic; and (3) Distributional Collapse, where signals fail to generalize without amplifying pre-training biases. To address these, we introduce LC-ERD (Logic-Consistent Endogenous Reward Decomposition), a framework framing self-alignment as latent structure mining. We derive a Variational Logic Potential by aggregating consensus from the model's Latent Logic Expertise (LLE) to denoise the reasoning manifold, and introduce a Multi-Agent Value Decomposition protocol based on the IGM principle to quantify individual step utility. Experiments show LC-ERD delivers a robust self-evolution path, uncovering trade-offs between logic consistency and accuracy while identifying high-value reasoning patterns missed by standard rewards. Our code is available at https://github.com/Reinhardmannn/LC-ERD.

Dianzhi Yu, Vireo Zhang, Hongru Wang et al.

Achieving self-evolution in intelligent agents requires the continual accumulation of new knowledge across changing task sequences without forgetting previously acquired abilities. Existing approaches either internalize knowledge by updating model parameters, which induces catastrophic forgetting, or rely on external memory, which fails to genuinely enhance the model's intrinsic capabilities. We propose MoLEM, a generative mixture of latent memory framework based on a dynamic mixture-of-experts (MoE). We treat multiple experts as independent carriers to generate memory. A router selects and weights experts through key-query matching, and the aggregated latent memory is injected into the reasoning process. The base model for reasoning remains entirely frozen, with all experiential knowledge internalized into the additional modules, avoiding catastrophic forgetting. For continual learning, each training stage is paired with a lightweight autoencoder that selects the appropriate routing group at inference, and inputs that match no stage fall back to the pretrained model. Experiments train the framework on continual-learning sequences spanning math, science, and code domains. After training, we evaluate the framework on the corresponding test sets to measure task learning and competence preservation across continual adaptation stages. After the full continual-learning sequence, our method improves the average accuracy by 10.40% over the Vanilla pretrained baseline, while none of the competing methods consistently exceed this baseline across different training orders.

Yi-Jun Chang, Yanyu Chen, Dipan Dey et al.

We investigate the \emph{minimum weight cycle (MWC)} problem in the $\mathsf{CONGEST}$ model of distributed computing. For undirected weighted graphs, we design a randomized algorithm that achieves a $(k+1)$-approximation, for any \emph{real} number $k \ge 1$. The round complexity of algorithm is \[ \tilde{O}\!\Big( n^{\frac{k+1}{2k+1}} + n^{\frac{1}{k}} + D\, n^{\frac{1}{2(2k+1)}} + D^{\frac{2}{5}} n^{\frac{2}{5}+\frac{1}{2(2k+1)}} \Big). \] where $n$ denotes the number of nodes and $D$ is the unweighted diameter of the graph. This result yields a smooth trade-off between approximation ratio and round complexity. In particular, when $k \geq 2$ and $D = \tilde{O}(n^{1/4})$, the bound simplifies to \[ \tilde{O}\!\left( n^{\frac{k+1}{2k+1}} \right) \] On the lower bound side, assuming the Erdős girth conjecture, we prove that for every \emph{integer} $k \ge 1$, any randomized $(k+1-ε)$-approximation algorithm for MWC requires \[ \tildeΩ\!\left( n^{\frac{k+1}{2k+1}} \right) \] rounds. This lower bound holds for both directed unweighted and undirected weighted graphs, and applies even to graphs with small diameter $D = Θ(\log n)$. Taken together, our upper and lower bounds \emph{match up to polylogarithmic factors} for graphs of sufficiently small diameter $D = \tilde{O}(n^{1/4})$ (when $k \geq 2$), yielding a nearly tight bound on the distributed complexity of the problem. Our results improve upon the previous state of the art: Manoharan and Ramachandran (PODC~2024) demonstrated a $(2+ε)$-approximation algorithm for undirected weighted graphs with round complexity $\tilde{O}(n^{2/3}+D)$, and proved that for any arbitrarily large number $α$, any $α$-approximation algorithm for directed unweighted or undirected weighted graphs requires $Ω(\sqrt{n}/\log n)$ rounds.

Yanyu Chen, Jiyue Jiang, Jiahong Liu et al.

The evaluation of Deep Research Agents is a critical challenge, as conventional outcome-based metrics fail to capture the nuances of their complex reasoning. Current evaluation faces two primary challenges: 1) a reliance on singular metrics like Pass@1, creating a "high-score illusion" that ignores the quality, efficiency, and soundness of the reasoning process; and 2) the failure of static benchmarks to quantify crucial attributes like robustness and latent capability. To address these gaps, we introduce TRACE (Trajectory-Aware Comprehensive Evaluation), a framework that holistically assesses the entire problem-solving trajectory. To counter the "high-score illusion", we propose a Hierarchical Trajectory Utility Function that quantifies process efficiency and cognitive quality, including evidence grounding, alongside accuracy. To measure deeper attributes, TRACE introduces a Scaffolded Capability Assessment protocol, quantifying an agent's latent ability by determining the minimum guidance needed for success. Our contributions include the TRACE framework, its novel metrics, and the accompanying DeepResearch-Bench with controllable complexity. Experiments show TRACE delivers a granular ranking that uncovers critical trade-offs between agent accuracy, efficiency, and robustness entirely missed by singular metrics.

Wenli Yang, Guan Huang, Renjie Li et al.

Convolutional neural network (CNN) models have seen advanced improvements in performance in various domains, but lack of interpretability is a major barrier to assurance and regulation during operation for acceptance and deployment of AI-assisted applications. There have been many works on input interpretability focusing on analyzing the input-output relations, but the internal logic of models has not been clarified in the current mainstream interpretability methods. In this study, we propose a novel hybrid CNN-interpreter through: (1) An original forward propagation mechanism to examine the layer-specific prediction results for local interpretability. (2) A new global interpretability that indicates the feature correlation and filter importance effects. By combining the local and global interpretabilities, hybrid CNN-interpreter enables us to have a solid understanding and monitoring of model context during the whole learning process with detailed and consistent representations. Finally, the proposed interpretabilities have been demonstrated to adapt to various CNN-based model structures.

Jiyue Jiang, Alfred Kar Yin Truong, Yanyu Chen et al.

High-quality data resources play a crucial role in learning large language models (LLMs), particularly for low-resource languages like Cantonese. Despite having more than 85 million native speakers, Cantonese is still considered a low-resource language in the field of natural language processing (NLP) due to factors such as the dominance of Mandarin, lack of cohesion within the Cantonese-speaking community, diversity in character encoding and input methods, and the tendency of overseas Cantonese speakers to prefer using English. In addition, rich colloquial vocabulary of Cantonese, English loanwords, and code-switching characteristics add to the complexity of corpus collection and processing. To address these challenges, we collect Cantonese texts from a variety of sources, including open source corpora, Hong Kong-specific forums, Wikipedia, and Common Crawl data. We conduct rigorous data processing through language filtering, quality filtering, content filtering, and de-duplication steps, successfully constructing a high-quality Cantonese corpus of over 2 billion tokens for training large language models. We further refined the model through supervised fine-tuning (SFT) on curated Cantonese tasks, enhancing its ability to handle specific applications. Upon completion of the training, the model achieves state-of-the-art (SOTA) performance on four Cantonese benchmarks. After training on our dataset, the model also exhibits improved performance on other mainstream language tasks.

Yanyu Chen, Yao Yao, Wai Kin Victor Chan et al.

Data sparsity and cold-start problems are persistent challenges in recommendation systems. Cross-domain recommendation (CDR) is a promising solution that utilizes knowledge from the source domain to improve the recommendation performance in the target domain. Previous CDR approaches have mainly followed the Embedding and Mapping (EMCDR) framework, which involves learning a mapping function to facilitate knowledge transfer. However, these approaches necessitate re-engineering and re-training the network structure to incorporate transferrable knowledge, which can be computationally expensive and may result in catastrophic forgetting of the original knowledge. In this paper, we present a scalable and efficient paradigm to address data sparsity and cold-start issues in CDR, named CDR-Adapter, by decoupling the original recommendation model from the mapping function, without requiring re-engineering the network structure. Specifically, CDR-Adapter is a novel plug-and-play module that employs adapter modules to align feature representations, allowing for flexible knowledge transfer across different domains and efficient fine-tuning with minimal training costs. We conducted extensive experiments on the benchmark dataset, which demonstrated the effectiveness of our approach over several state-of-the-art CDR approaches.

Hongyi Zheng, Yanyu Chen, Zihan Zhang

Our project probes the relationship between temperatures and the blossom date of cherry trees. Through modeling, future flowering will become predictive, helping the public plan travels and avoid pollen season. To predict the date when the cherry trees will blossom exactly could be viewed as a multiclass classification problem, so we applied the multi-class Support Vector Classifier (SVC) and Recurrent Neural Network (RNN), particularly Long Short-term Memory (LSTM), to formulate the problem. In the end, we evaluate and compare the performance of these approaches to find out which one might be more applicable in reality.

Jiahang Zhou, Yanyu Chen, Zicong Hong et al.

Foundation models (e.g., ChatGPT, DALL-E, PengCheng Mind, PanGu-$Σ$) have demonstrated extraordinary performance in key technological areas, such as natural language processing and visual recognition, and have become the mainstream trend of artificial general intelligence. This has led more and more major technology giants to dedicate significant human and financial resources to actively develop their foundation model systems, which drives continuous growth of these models' parameters. As a result, the training and serving of these models have posed significant challenges, including substantial computing power, memory consumption, bandwidth demands, etc. Therefore, employing efficient training and serving strategies becomes particularly crucial. Many researchers have actively explored and proposed effective methods. So, a comprehensive survey of them is essential for system developers and researchers. This paper extensively explores the methods employed in training and serving foundation models from various perspectives. It provides a detailed categorization of these state-of-the-art methods, including finer aspects such as network, computing, and storage. Additionally, the paper summarizes the challenges and presents a perspective on the future development direction of foundation model systems. Through comprehensive discussion and analysis, it hopes to provide a solid theoretical basis and practical guidance for future research and applications, promoting continuous innovation and development in foundation model systems.

Jiyue Jiang, Pengan Chen, Jiuming Wang et al.

Large language models (LLMs) have become important tools in solving biological problems, offering improvements in accuracy and adaptability over conventional methods. Several benchmarks have been proposed to evaluate the performance of these LLMs. However, current benchmarks can hardly evaluate the performance of these models across diverse tasks effectively. In this paper, we introduce a comprehensive prompting-based benchmarking framework, termed Bio-benchmark, which includes 30 key bioinformatics tasks covering areas such as proteins, RNA, drugs, electronic health records, and traditional Chinese medicine. Using this benchmark, we evaluate six mainstream LLMs, including GPT-4o and Llama-3.1-70b, etc., using 0-shot and few-shot Chain-of-Thought (CoT) settings without fine-tuning to reveal their intrinsic capabilities. To improve the efficiency of our evaluations, we demonstrate BioFinder, a new tool for extracting answers from LLM responses, which increases extraction accuracy by round 30% compared to existing methods. Our benchmark results show the biological tasks suitable for current LLMs and identify specific areas requiring enhancement. Furthermore, we propose targeted prompt engineering strategies for optimizing LLM performance in these contexts. Based on these findings, we provide recommendations for the development of more robust LLMs tailored for various biological applications. This work offers a comprehensive evaluation framework and robust tools to support the application of LLMs in bioinformatics.

Yanting Li, Jiyue Jiang, Zikang Wang et al.

Inverse Protein Folding (IPF) is a critical subtask in the field of protein design, aiming to engineer amino acid sequences capable of folding correctly into a specified three-dimensional (3D) conformation. Although substantial progress has been achieved in recent years, existing methods generally rely on either backbone coordinates or molecular surface features alone, which restricts their ability to fully capture the complex chemical and geometric constraints necessary for precise sequence prediction. To address this limitation, we present DS-ProGen, a dual-structure deep language model for functional protein design, which integrates both backbone geometry and surface-level representations. By incorporating backbone coordinates as well as surface chemical and geometric descriptors into a next-amino-acid prediction paradigm, DS-ProGen is able to generate functionally relevant and structurally stable sequences while satisfying both global and local conformational constraints. On the PRIDE dataset, DS-ProGen attains the current state-of-the-art recovery rate of 61.47%, demonstrating the synergistic advantage of multi-modal structural encoding in protein design. Furthermore, DS-ProGen excels in predicting interactions with a variety of biological partners, including ligands, ions, and RNA, confirming its robust functional retention capabilities.

Wenli Yang, Yanyu Chen, Andrew Trotter et al.

Phenotype segmentation is pivotal in analysing visual features of living organisms, enhancing our understanding of their characteristics. In the context of oysters, meat quality assessment is paramount, focusing on shell, meat, gonad, and muscle components. Traditional manual inspection methods are time-consuming and subjective, prompting the adoption of machine vision technology for efficient and objective evaluation. We explore machine vision's capacity for segmenting oyster components, leading to the development of a multi-network ensemble approach with a global-local hierarchical attention mechanism. This approach integrates predictions from diverse models and addresses challenges posed by varying scales, ensuring robust instance segmentation across components. Finally, we provide a comprehensive evaluation of the proposed method's performance using different real-world datasets, highlighting its efficacy and robustness in enhancing oyster phenotype segmentation.

Yanyu Chen, Ganhong Huang

Efficiently deploying large language models (LLMs) in real-world scenarios remains a critical challenge, primarily due to hardware heterogeneity, inference framework limitations, and workload complexities.Efficiently deploying large language models (LLMs) in real-world scenarios remains a critical challenge, primarily due to hardware heterogeneity, inference framework limitations, and workload complexities. These challenges often lead to inefficiencies in memory utilization, latency, and throughput, hindering the effective deployment of LLMs, especially for non-experts. Through extensive experiments, we identify key performance bottlenecks, including sudden drops in memory utilization, latency fluctuations with varying batch sizes, and inefficiencies in multi-GPU configurations. These insights reveal a vast optimization space shaped by the intricate interplay of hardware, frameworks, and workload parameters. This underscores the need for a systematic approach to optimize LLM inference, motivating the design of our framework, GUIDE. GUIDE leverages dynamic modeling and simulation-based optimization to address these issues, achieving prediction errors between 9.9% and 42.3% for key metrics such as batch latency, TTFT, and decode throughput. By effectively bridging the gap between theoretical performance and practical deployment, our framework empowers practitioners, particularly non-specialists, to make data-driven decisions and unlock the full potential of LLMs in heterogeneous environments cheaply.