Cheng Li

Cheng Li, Mingyang Zhang, Qiaozhu Mei et al. · baidu, deepmind

Personalized text generation is an emerging research area that has attracted much attention in recent years. Most studies in this direction focus on a particular domain by designing bespoke features or models. In this work, we propose a general approach for personalized text generation using large language models (LLMs). Inspired by the practice of writing education, we develop a multistage and multitask framework to teach LLMs for personalized generation. In writing instruction, the task of writing from sources is often decomposed into multiple steps that involve finding, evaluating, summarizing, synthesizing, and integrating information. Analogously, our approach to personalized text generation consists of multiple stages: retrieval, ranking, summarization, synthesis, and generation. In addition, we introduce a multitask setting that helps the model improve its generation ability further, which is inspired by the observation in education that a student's reading proficiency and writing ability are often correlated. We evaluate our approach on three public datasets, each of which covers a different and representative domain. Our results show significant improvements over a variety of baselines.

Kimi Team, Tongtong Bai, Yifan Bai et al.

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

Weijian Huang, Cheng Li, Hong-Yu Zhou et al.

Recently, multi-modal vision-language foundation models have gained significant attention in the medical field. While these models offer great opportunities, they still face crucial challenges, such as the requirement for fine-grained knowledge understanding in computer-aided diagnosis and the capability of utilizing very limited or even no task-specific labeled data in real-world clinical applications. In this study, we present MaCo, a masked contrastive chest X-ray foundation model that tackles these challenges. MaCo explores masked contrastive learning to simultaneously achieve fine-grained image understanding and zero-shot learning for a variety of medical imaging tasks. It designs a correlation weighting mechanism to adjust the correlation between masked chest X-ray image patches and their corresponding reports, thereby enhancing the model's representation learning capabilities. To evaluate the performance of MaCo, we conducted extensive experiments using 6 well-known open-source X-ray datasets. The experimental results demonstrate the superiority of MaCo over 10 state-of-the-art approaches across tasks such as classification, segmentation, detection, and phrase grounding. These findings highlight the significant potential of MaCo in advancing a wide range of medical image analysis tasks.

George Ciubotariu, Sharif S M A, Abdur Rehman et al.

This paper presents a comprehensive review of the NTIRE 2026 Low Light Image Enhancement Challenge, highlighting the proposed solutions and final results. The objective of this challenge is to identify effective networks capable of producing clearer and visually compelling images in diverse and challenging conditions by learning representative visual cues with the purpose of restoring information loss due to low-contrast and noisy images. A total of 195 participants registered for the first track and 153 for the second track of the competition, and 22 teams ultimately submitted valid entries. This paper thoroughly evaluates the state-of-the-art advances in (joint denoising and) low-light image enhancement, showcasing the significant progress in the field, while leveraging samples of our novel dataset.

Zhiqi Lin, Youshan Miao, Guanbin Xu et al. · microsoft-research

Increasingly complex and diverse deep neural network (DNN) models necessitate distributing the execution across multiple devices for training and inference tasks, and also require carefully planned schedules for performance. However, existing practices often rely on predefined schedules that may not fully exploit the benefits of emerging diverse model-aware operator placement strategies. Handcrafting high-efficiency schedules can be challenging due to the large and varying schedule space. This paper presents Tessel, an automated system that searches for efficient schedules for distributed DNN training and inference for diverse operator placement strategies. To reduce search costs, Tessel leverages the insight that the most efficient schedules often exhibit repetitive pattern (repetend) across different data inputs. This leads to a two-phase approach: repetend construction and schedule completion. By exploring schedules for various operator placement strategies, Tessel significantly improves both training and inference performance. Experiments with representative DNN models demonstrate that Tessel achieves up to 5.5x training performance speedup and up to 38% inference latency reduction.

Zhiqi Lin, Youshan Miao, Guodong Liu et al. · microsoft-research

With the growing model size, deep neural networks (DNN) are increasingly trained over massive GPU accelerators, which demands a proper parallelization plan that transforms a DNN model into fine-grained tasks and then schedules them to GPUs for execution. Due to the large search space, the contemporary parallelization plan generators often rely on empirical rules that couple transformation and scheduling, and fall short in exploring more flexible schedules that yield better memory usage and compute efficiency. This tension can be exacerbated by the emerging models with increasing complexity in their structure and model size. SuperScaler is a system that facilitates the design and generation of highly flexible parallelization plans. It formulates the plan design and generation into three sequential phases explicitly: model transformation, space-time scheduling, and data dependency preserving. Such a principled approach decouples multiple seemingly intertwined factors and enables the composition of highly flexible parallelization plans. As a result, SuperScaler can not only generate empirical parallelization plans, but also construct new plans that achieve up to 3.5X speedup compared to state-of-the-art solutions like DeepSpeed, Megatron and Alpa, for emerging DNN models like Swin-Transformer and AlphaFold2, as well as well-optimized models like GPT-3.

Weijian Huang, Hao Yang, Cheng Li et al.

Multi-modal representation methods have achieved advanced performance in medical applications by extracting more robust features from multi-domain data. However, existing methods usually need to train additional branches for downstream tasks, which may increase the model complexities in clinical applications as well as introduce additional human inductive bias. Besides, very few studies exploit the rich clinical knowledge embedded in clinical daily reports. To this end, we propose a novel medical generalist agent, MGA, that can address three kinds of common clinical tasks via clinical reports knowledge transformation. Unlike the existing methods, MGA can easily adapt to different tasks without specific downstream branches when their corresponding annotations are missing. More importantly, we are the first attempt to use medical professional language guidance as a transmission medium to guide the agent's behavior. The proposed method is implemented on four well-known X-ray open-source datasets, MIMIC-CXR, CheXpert, MIMIC-CXR-JPG, and MIMIC-CXR-MS. Promising results are obtained, which validate the effectiveness of our proposed MGA. Code is available at: https://github.com/SZUHvern/MGA

Yizun Lin, Jian-Feng Cai, Zhao-Rong Lai et al.

This paper investigates a category of constrained fractional optimization problems that emerge in various practical applications. The objective function for this category is characterized by the ratio of a numerator and denominator, both being convex, semi-algebraic, Lipschitz continuous, and differentiable with Lipschitz continuous gradients over the constraint sets. The constrained sets associated with these problems are closed, convex, and semi-algebraic. We propose an efficient algorithm that is inspired by the proximal gradient method, and we provide a thorough convergence analysis. Our algorithm offers several benefits compared to existing methods. It requires only a single proximal gradient operation per iteration, thus avoiding the complicated inner-loop concave maximization usually required. Additionally, our method converges to a critical point without the typical need for a nonnegative numerator, and this critical point becomes a globally optimal solution with an appropriate condition. Our approach is adaptable to unbounded constraint sets as well. Therefore, our approach is viable for many more practical models. Numerical experiments show that our method not only reliably reaches ground-truth solutions in some model problems but also outperforms several existing methods in maximizing the Sharpe ratio with real-world financial data.

Yifan Song, Weimin Xiong, Dawei Zhu et al. · pku

Tool-augmented large language models (LLMs) have achieved remarkable progress in tackling a broad range of tasks. However, existing methods are mainly restricted to specifically designed tools and fail to fulfill complex instructions, having great limitations when confronted with real-world scenarios. In this paper, we explore a more realistic scenario by connecting LLMs with RESTful APIs, which adhere to the widely adopted REST software architectural style for web service development. To address the practical challenges of tackling complex instructions, we propose RestGPT, which exploits the power of LLMs and conducts a coarse-to-fine online planning mechanism to enhance the abilities of task decomposition and API selection. RestGPT also contains an API executor tailored for calling RESTful APIs, which can meticulously formulate parameters and parse API responses. To fully evaluate the performance of RestGPT, we propose RestBench, a high-quality benchmark which consists of two real-world scenarios and human-annotated instructions with gold solution paths. Experiments show that RestGPT is able to achieve impressive results in complex tasks and has strong robustness, which paves a new way towards AGI. RestGPT and RestBench is publicly available at https://restgpt.github.io/.

Cheng Li, Jindong Wang, Yixuan Zhang et al.

Emotional intelligence significantly impacts our daily behaviors and interactions. Although Large Language Models (LLMs) are increasingly viewed as a stride toward artificial general intelligence, exhibiting impressive performance in numerous tasks, it is still uncertain if LLMs can genuinely grasp psychological emotional stimuli. Understanding and responding to emotional cues gives humans a distinct advantage in problem-solving. In this paper, we take the first step towards exploring the ability of LLMs to understand emotional stimuli. To this end, we first conduct automatic experiments on 45 tasks using various LLMs, including Flan-T5-Large, Vicuna, Llama 2, BLOOM, ChatGPT, and GPT-4. Our tasks span deterministic and generative applications that represent comprehensive evaluation scenarios. Our automatic experiments show that LLMs have a grasp of emotional intelligence, and their performance can be improved with emotional prompts (which we call "EmotionPrompt" that combines the original prompt with emotional stimuli), e.g., 8.00% relative performance improvement in Instruction Induction and 115% in BIG-Bench. In addition to those deterministic tasks that can be automatically evaluated using existing metrics, we conducted a human study with 106 participants to assess the quality of generative tasks using both vanilla and emotional prompts. Our human study results demonstrate that EmotionPrompt significantly boosts the performance of generative tasks (10.9% average improvement in terms of performance, truthfulness, and responsibility metrics). We provide an in-depth discussion regarding why EmotionPrompt works for LLMs and the factors that may influence its performance. We posit that EmotionPrompt heralds a novel avenue for exploring interdisciplinary knowledge for human-LLMs interaction.

Marcos V. Conde, Radu Timofte, Yibin Huang et al.

Cameras capture sensor RAW images and transform them into pleasant RGB images, suitable for the human eyes, using their integrated Image Signal Processor (ISP). Numerous low-level vision tasks operate in the RAW domain (e.g. image denoising, white balance) due to its linear relationship with the scene irradiance, wide-range of information at 12bits, and sensor designs. Despite this, RAW image datasets are scarce and more expensive to collect than the already large and public RGB datasets. This paper introduces the AIM 2022 Challenge on Reversed Image Signal Processing and RAW Reconstruction. We aim to recover raw sensor images from the corresponding RGBs without metadata and, by doing this, "reverse" the ISP transformation. The proposed methods and benchmark establish the state-of-the-art for this low-level vision inverse problem, and generating realistic raw sensor readings can potentially benefit other tasks such as denoising and super-resolution.

Cheng Li, Ziang Leng, Chenxi Yan et al.

Role-playing chatbots built on large language models have drawn interest, but better techniques are needed to enable mimicking specific fictional characters. We propose an algorithm that controls language models via an improved prompt and memories of the character extracted from scripts. We construct ChatHaruhi, a dataset covering 32 Chinese / English TV / anime characters with over 54k simulated dialogues. Both automatic and human evaluations show our approach improves role-playing ability over baselines. Code and data are available at https://github.com/LC1332/Chat-Haruhi-Suzumiya .

Reza Yazdani Aminabadi, Samyam Rajbhandari, Minjia Zhang et al.

The past several years have witnessed the success of transformer-based models, and their scale and application scenarios continue to grow aggressively. The current landscape of transformer models is increasingly diverse: the model size varies drastically with the largest being of hundred-billion parameters; the model characteristics differ due to the sparsity introduced by the Mixture-of-Experts; the target application scenarios can be latency-critical or throughput-oriented; the deployment hardware could be single- or multi-GPU systems with different types of memory and storage, etc. With such increasing diversity and the fast-evolving pace of transformer models, designing a highly performant and efficient inference system is extremely challenging. In this paper, we present DeepSpeed Inference, a comprehensive system solution for transformer model inference to address the above-mentioned challenges. DeepSpeed Inference consists of (1) a multi-GPU inference solution to minimize latency while maximizing the throughput of both dense and sparse transformer models when they fit in aggregate GPU memory, and (2) a heterogeneous inference solution that leverages CPU and NVMe memory in addition to the GPU memory and compute to enable high inference throughput with large models which do not fit in aggregate GPU memory. DeepSpeed Inference reduces latency by up to 7.3X over the state-of-the-art for latency-oriented scenarios and increases throughput by over 1.5x for throughput-oriented scenarios. Moreover, it enables trillion parameter scale inference under real-time latency constraints by leveraging hundreds of GPUs, an unprecedented scale for inference. It can inference 25x larger models than with GPU-only solutions, while delivering a high throughput of 84 TFLOPS (over $50\%$ of A6000 peak).

Weijian Huang, Cheng Li, Wenxin Fan et al.

Recovering high-quality images from undersampled measurements is critical for accelerated MRI reconstruction. Recently, various supervised deep learning-based MRI reconstruction methods have been developed. Despite the achieved promising performances, these methods require fully sampled reference data, the acquisition of which is resource-intensive and time-consuming. Self-supervised learning has emerged as a promising solution to alleviate the reliance on fully sampled datasets. However, existing self-supervised methods suffer from reconstruction errors due to the insufficient constraint enforced on the non-sampled data points and the error accumulation happened alongside the iterative image reconstruction process for model-driven deep learning reconstrutions. To address these challenges, we propose K2Calibrate, a K-space adaptation strategy for self-supervised model-driven MR reconstruction optimization. By iteratively calibrating the learned measurements, K2Calibrate can reduce the network's reconstruction deterioration caused by statistically dependent noise. Extensive experiments have been conducted on the open-source dataset FastMRI, and K2Calibrate achieves better results than five state-of-the-art methods. The proposed K2Calibrate is plug-and-play and can be easily integrated with different model-driven deep learning reconstruction methods.

Zhewei Yao, Xiaoxia Wu, Cheng Li et al.

Post-training quantization (PTQ) has emerged as a promising technique for mitigating memory consumption and computational costs in large language models (LLMs). However, a systematic examination of various quantization schemes, model families, and quantization bit precision has been absent from the literature. In this paper, we conduct a comprehensive analysis of these factors by investigating the effects of PTQ on weight-only, activation-only, and weight-and-activation quantization using diverse methods such as round-to-nearest (RTN), GPTQ, ZeroQuant, and their variants. We apply these methods to two distinct model families with parameters ranging from 125M to 176B. Our contributions include: (1) a sensitivity analysis revealing that activation quantization is generally more susceptible to weight quantization, with smaller models often outperforming larger models in terms of activation quantization; (2) an evaluation and comparison of existing PTQ methods to optimize model size reduction while minimizing the impact on accuracy, revealing that none of the current methods can achieve the original model quality for quantization with either INT4-weight or INT4-weight-and-INT8-activation; (3) based on these insights, we propose an optimized method called Low-Rank Compensation (LoRC), which employs low-rank matrices to enhance model quality recovery with a minimal increase in model size.

Conglong Li, Zhewei Yao, Xiaoxia Wu et al.

Recent advances on deep learning models come at the price of formidable training cost. The increasing model size is one of the root causes, but another less-emphasized fact is that data scale is actually increasing at a similar speed as model scale, and the training cost is proportional to both of them. Compared to the rapidly evolving model architecture, how to efficiently use the training data (especially for the expensive foundation model pretraining) is both less explored and difficult to realize due to the lack of a convenient framework that focuses on data efficiency capabilities. To this end, we present DeepSpeed Data Efficiency, a framework that makes better use of data, increases training efficiency, and improves model quality. Specifically, we propose and combine two data efficiency techniques: efficient data sampling via a general curriculum learning library, and efficient data routing via a novel random layerwise token dropping technique. For GPT-3 1.3B language model pretraining, our work achieves 12.5x less data/time/cost (\$3.7K if rent on Azure), while still maintaining 95% of model quality compared to baseline with full data and cost (\$46.3K). For GPT-3 1.3B and BERT-large pretraining, our work can also achieve the same model quality with up to 2x less data/time/cost, or achieve better model quality under same data/time/cost. DeepSpeed Data Efficiency is easy to use and tune, enabling us to easily apply it and verify its benefit on additional tasks including GPT-3 MoE model pretraining and small-scale GPT-2/ViT finetuning.

Xintao Wang, Yunze Xiao, Jen-tse Huang et al.

Role-playing agents (RPAs), powered by large language models, have emerged as a flourishing field of applications. However, a key challenge lies in assessing whether RPAs accurately reproduce the personas of target characters, namely their character fidelity. Existing methods mainly focus on the knowledge and linguistic patterns of characters. This paper, instead, introduces a novel perspective to evaluate the personality fidelity of RPAs with psychological scales. Overcoming drawbacks of previous self-report assessments on RPAs, we propose InCharacter, namely Interviewing Character agents for personality tests. Experiments include various types of RPAs and LLMs, covering 32 distinct characters on 14 widely used psychological scales. The results validate the effectiveness of InCharacter in measuring RPA personalities. Then, with InCharacter, we show that state-of-the-art RPAs exhibit personalities highly aligned with the human-perceived personalities of the characters, achieving an accuracy up to 80.7%.

Duanchu Wang, Cheng Li, Junjie Yang et al.

Point cloud quality plays a critical role in 3D acquisition, reconstruction, rendering, and perception, yet existing point cloud quality assessment (PCQA) research remains largely centered on scalar score prediction. In practical inspection scenarios, quality assessment often involves identifying defects, characterizing dominant issue types, assessing downstream usability, and providing evidence-supported descriptions, which are not explicitly evaluated by current benchmarks. We introduce PointQ-Bench, a benchmark designed to extend PCQA from scalar scoring toward comprehensive quality understanding. PointQ-Bench consists of 3,083 point clouds spanning authentic scans, simulated distortions, and AI-generated content, covering eight major issue types. Each sample is annotated with mean opinion scores (MOS), quality levels, issue tags, expert-grounded descriptions, and 12,332 question-answer pairs. The benchmark supports three perception-oriented tasks: anomaly sensing, defect diagnosis, and usability grading, as well as a cognition-oriented task of open-ended quality reporting. To evaluate free-form quality descriptions, we further propose SSFRQ-5D, a five-dimensional evaluation protocol validated through human-AI agreement analysis. Extensive experiments on 14 vision-language models and traditional PCQA baselines reveal a consistent perception-diagnosis gap: while current models exhibit emerging abilities in coarse defect perception, they struggle with grounded diagnosis and quality calibration. Strong 2D MLLMs generally outperform existing 3D VLMs, and the benefit of additional views or point-level inputs is non-uniform, varying across tasks, data sources, and models, particularly under boundary-ambiguous conditions. Overall, PointQ-Bench provides a diagnostic testbed for advancing reliable and interpretable point cloud quality understanding.

Ziyin Zhang, Ning Lu, Minghui Liao et al.

Text recognition methods are gaining rapid development. Some advanced techniques, e.g., powerful modules, language models, and un- and semi-supervised learning schemes, consecutively push the performance on public benchmarks forward. However, the problem of how to better optimize a text recognition model from the perspective of loss functions is largely overlooked. CTC-based methods, widely used in practice due to their good balance between performance and inference speed, still grapple with accuracy degradation. This is because CTC loss emphasizes the optimization of the entire sequence target while neglecting to learn individual characters. We propose a self-distillation scheme for CTC-based model to address this issue. It incorporates a framewise regularization term in CTC loss to emphasize individual supervision, and leverages the maximizing-a-posteriori of latent alignment to solve the inconsistency problem that arises in distillation between CTC-based models. We refer to the regularized CTC loss as Distillation Connectionist Temporal Classification (DCTC) loss. DCTC loss is module-free, requiring no extra parameters, longer inference lag, or additional training data or phases. Extensive experiments on public benchmarks demonstrate that DCTC can boost text recognition model accuracy by up to 2.6%, without any of these drawbacks.

Xiaoxia Wu, Cheng Li, Reza Yazdani Aminabadi et al.

Improving the deployment efficiency of transformer-based language models has been challenging given their high computation and memory cost. While INT8 quantization has recently been shown to be effective in reducing both the memory cost and latency while preserving model accuracy, it remains unclear whether we can leverage INT4 (which doubles peak hardware throughput) to achieve further latency improvement. In this study, we explore the feasibility of employing INT4 weight and activation (W4A4) quantization for language models. Our findings indicate that W4A4 quantization introduces no to negligible accuracy degradation for encoder-only and encoder-decoder models, but causes a significant accuracy drop for decoder-only models. To materialize the performance gain using W4A4, we develop a highly optimized end-to-end W4A4 encoder inference pipeline supporting different quantization strategies. Our INT4 pipeline is $8.5\times$ faster for latency-oriented scenarios and up to $3\times$ for throughput-oriented scenarios compared to the inference of FP16, and improves the SOTA BERT INT8 performance from FasterTransformer by up to $1.7\times$. We provide insights into the failure cases when applying W4A4 to decoder-only models, and further explore the compatibility of INT4 quantization with other compression methods, like pruning and layer reduction.

Hao Yang, Weijian Huang, Jiarun Liu et al.

The ability to incrementally learn new classes from limited samples is crucial to the development of artificial intelligence systems for real clinical application. Although existing incremental learning techniques have attempted to address this issue, they still struggle with only few labeled data, particularly when the samples are from varied domains. In this paper, we explore the cross-domain few-shot incremental learning (CDFSCIL) problem. CDFSCIL requires models to learn new classes from very few labeled samples incrementally, and the new classes may be vastly different from the target space. To counteract this difficulty, we propose a cross-domain enhancement constraint and cross-domain data augmentation method. Experiments on MedMNIST show that the classification performance of this method is better than other similar incremental learning methods.

Weixun Wang, XiaoXiao Xu, Wanhe An et al.

Agentic crafting requires LLMs to operate in real-world environments over multiple turns by taking actions, observing outcomes, and iteratively refining artifacts. Despite its importance, the open-source community lacks a principled, end-to-end ecosystem to streamline agent development. We introduce the Agentic Learning Ecosystem (ALE), a foundational infrastructure that optimizes the production pipeline for agentic model. ALE consists of three components: ROLL, a post-training framework for weight optimization; ROCK, a sandbox environment manager for trajectory generation; and iFlow CLI, an agent framework for efficient context engineering. We release ROME, an open-source agent grounded by ALE and trained on over one million trajectories. Our approach includes data composition protocols for synthesizing complex behaviors and a novel policy optimization algorithm, Interaction-Perceptive Agentic Policy Optimization (IPA), which assigns credit over semantic interaction chunks rather than individual tokens to improve long-horizon training stability. Empirically, we evaluate ROME within a structured setting and introduce Terminal Bench Pro, a benchmark with improved scale and contamination control. ROME demonstrates strong performance across benchmarks like SWE-bench Verified and Terminal Bench, proving the effectiveness of ALE.

Zihang Cheng, Duanchu Wang, Cheng Li et al.

Remote sensing question answering (RS-QA) often requires more than direct semantic prediction, especially in large-scale forest scenes where ecological analysis involves multi-step filtering, numerical aggregation, neighborhood reasoning, and verifiable evidence. We introduce ForestHG-Trace, a framework for traceable long-horizon ecological reasoning over forest environments. It represents multimodal NEON forest scenes as ecological hypergraphs, where tree instances, spatial units, semantic groups, and neighborhood relations support higher-order reasoning beyond pairwise scene graphs. An LLM-guided agent then invokes deterministic tools for reading, filtering, expansion, aggregation, comparison, and auditing, producing replayable execution traces and compact evidence records rather than only free-form answers. We further construct ForestTraceQA, an executable benchmark for evaluating ecological QA across diverse task types and reasoning depths. Experiments show that ForestHG-Trace substantially improves answer accuracy and execution faithfulness over single-step baselines and scene-graph agents, while highlighting execution depth as the main bottleneck for long-horizon ecological QA.

Wei Li, Shuai Xiao, Tianhong Dai et al.

Ghosting artifacts, motion blur, and low fidelity in highlight are the main challenges in High Dynamic Range (HDR) imaging from multiple Low Dynamic Range (LDR) images. These issues come from using the medium-exposed image as the reference frame in previous methods. To deal with them, we propose to use the under-exposed image as the reference to avoid these issues. However, the heavy noise in dark regions of the under-exposed image becomes a new problem. Therefore, we propose a joint HDR and denoising pipeline, containing two sub-networks: (i) a pre-denoising network (PreDNNet) to adaptively denoise input LDRs by exploiting exposure priors; (ii) a pyramid cascading fusion network (PCFNet), introducing an attention mechanism and cascading structure in a multi-scale manner. To further leverage these two paradigms, we propose a selective and joint HDR and denoising (SJ-HD$^2$R) imaging framework, utilizing scenario-specific priors to conduct the path selection with an accuracy of more than 93.3$\%$. We create the first joint HDR and denoising benchmark dataset, which contains a variety of challenging HDR and denoising scenes and supports the switching of the reference image. Extensive experiment results show that our method achieves superior performance to previous methods.

Zhewei Yao, Xiaoxia Wu, Conglong Li et al.

Large-scale transformer models have become the de-facto architectures for various machine learning applications, e.g., CV and NLP. However, those large models also introduce prohibitive training costs. To mitigate this issue, we propose a novel random and layerwise token dropping method (random-LTD), which skips the computation of a subset of the input tokens at all middle layers. Particularly, random-LTD achieves considerable speedups and comparable accuracy as the standard training baseline. Compared to other token dropping methods, random-LTD does not require (1) any importance score-based metrics, (2) any special token treatment (e.g., [CLS]), and (3) many layers in full sequence length training except the first and the last layers. Besides, a new LayerToken learning rate schedule is proposed for pretraining problems that resolve the heavy tuning requirement for our proposed training mechanism. Finally, we demonstrate that random-LTD can be applied to broader applications, including GPT and BERT pretraining as well as ViT and GPT finetuning tasks. Our results show that random-LTD can save about 33.3% theoretical compute cost and 25.6% wall-clock training time while achieving similar zero-shot evaluations on GPT-31.3B as compared to baseline.

Jing Huang, Duanchu Wang, Junjie Yang et al.

Multimodal Large Language Models (MLLMs) have recently shown promising progress in geospatial reasoning. However, existing remote sensing benchmarks remain largely 2D-centric, evaluating models primarily on optical appearance. In natural environments, this paradigm breaks down due to severe spectral confusion, where ecologically distinct regions share similar textures but differ fundamentally in vertical structure. In such cases, explicit 3D structural data, such as Canopy Height Models (CHMs), become essential geometric evidence for semantic disambiguation. Yet, it remains unclear whether current MLLMs can genuinely leverage vertical cues to resolve appearance-level ambiguity. To address this gap, we introduce VertiCue-Bench, the first diagnostic benchmark for CHM-grounded geospatial reasoning. VertiCue-Bench comprises 1,534 carefully curated instances across 17 tasks, explicitly disentangling low-level height perception from ambiguity-aware semantic reasoning. Evaluations on 14 state-of-the-art general and remote-sensing-specialized MLLMs, combined with counterfactual modality testing, reveal a striking perception-reasoning dissociation. While models exhibit emerging competence in reading raw CHM height cues, they largely fail to translate geometric perception into reliable semantic reasoning, often underperforming RGB-only baselines when joint constraints are required. Overall, VertiCue-Bench exposes a critical geometry-to-semantics gap in natural scene understanding, offering actionable insights for advancing geospatial MLLMs.

Chi Zhang, Xiang Zhang, Mingyuan Lin et al.

Even though the collaboration between traditional and neuromorphic event cameras brings prosperity to frame-event based vision applications, the performance is still confined by the resolution gap crossing two modalities in both spatial and temporal domains. This paper is devoted to bridging the gap by increasing the temporal resolution for images, i.e., motion deblurring, and the spatial resolution for events, i.e., event super-resolving, respectively. To this end, we introduce CrossZoom, a novel unified neural Network (CZ-Net) to jointly recover sharp latent sequences within the exposure period of a blurry input and the corresponding High-Resolution (HR) events. Specifically, we present a multi-scale blur-event fusion architecture that leverages the scale-variant properties and effectively fuses cross-modality information to achieve cross-enhancement. Attention-based adaptive enhancement and cross-interaction prediction modules are devised to alleviate the distortions inherent in Low-Resolution (LR) events and enhance the final results through the prior blur-event complementary information. Furthermore, we propose a new dataset containing HR sharp-blurry images and the corresponding HR-LR event streams to facilitate future research. Extensive qualitative and quantitative experiments on synthetic and real-world datasets demonstrate the effectiveness and robustness of the proposed method. Codes and datasets are released at https://bestrivenzc.github.io/CZ-Net/.

Yuxin Ma, Ping Gong, Jun Yi et al.

Graph Neural Networks (GNNs) is a promising approach for applications with nonEuclidean data. However, training GNNs on large scale graphs with hundreds of millions nodes is both resource and time consuming. Different from DNNs, GNNs usually have larger memory footprints, and thus the GPU memory capacity and PCIe bandwidth are the main resource bottlenecks in GNN training. To address this problem, we present BiFeat: a graph feature quantization methodology to accelerate GNN training by significantly reducing the memory footprint and PCIe bandwidth requirement so that GNNs can take full advantage of GPU computing capabilities. Our key insight is that unlike DNN, GNN is less prone to the information loss of input features caused by quantization. We identify the main accuracy impact factors in graph feature quantization and theoretically prove that BiFeat training converges to a network where the loss is within $ε$ of the optimal loss of uncompressed network. We perform extensive evaluation of BiFeat using several popular GNN models and datasets, including GraphSAGE on MAG240M, the largest public graph dataset. The results demonstrate that BiFeat achieves a compression ratio of more than 30 and improves GNN training speed by 200%-320% with marginal accuracy loss. In particular, BiFeat achieves a record by training GraphSAGE on MAG240M within one hour using only four GPUs.

Kwok P Chun, Thanti Octavianti, Nilay Dogulu et al.

Managing complex disaster risks requires interdisciplinary efforts. Breaking down silos between law, social sciences, and natural sciences is critical for all processes of disaster risk reduction. This enables adaptive systems for the rapid evolution of AI technology, which has significantly impacted the intersection of law and natural environments. Exploring how AI influences legal frameworks and environmental management, while also examining how legal and environmental considerations can confine AI within the socioeconomic domain, is essential. From a co-production review perspective, drawing on insights from lawyers, social scientists, and environmental scientists, principles for responsible data mining are proposed based on safety, transparency, fairness, accountability, and contestability. This discussion offers a blueprint for interdisciplinary collaboration to create adaptive law systems based on AI integration of knowledge from environmental and social sciences. Discrepancies in the use of language between environmental scientists and decision-makers in terms of usefulness and accuracy hamper how AI can be used based on the principles of legal considerations for a safe, trustworthy, and contestable disaster management framework. When social networks are useful for mitigating disaster risks based on AI, the legal implications related to privacy and liability of the outcomes of disaster management must be considered. Fair and accountable principles emphasise environmental considerations and foster socioeconomic discussions related to public engagement. AI also has an important role to play in education, bringing together the next generations of law, social sciences, and natural sciences to work on interdisciplinary solutions in harmony.

Juan Zou, Cheng Li, Sen Jia et al.

Lately, deep learning has been extensively investigated for accelerating dynamic magnetic resonance (MR) imaging, with encouraging progresses achieved. However, without fully sampled reference data for training, current approaches may have limited abilities in recovering fine details or structures. To address this challenge, this paper proposes a self-supervised collaborative learning framework (SelfCoLearn) for accurate dynamic MR image reconstruction from undersampled k-space data. The proposed framework is equipped with three important components, namely, dual-network collaborative learning, reunderampling data augmentation and a specially designed co-training loss. The framework is flexible to be integrated with both data-driven networks and model-based iterative un-rolled networks. Our method has been evaluated on in-vivo dataset and compared it to four state-of-the-art methods. Results show that our method possesses strong capabilities in capturing essential and inherent representations for direct reconstructions from the undersampled k-space data and thus enables high-quality and fast dynamic MR imaging.

Zhuowan Li, Cheng Li, Mingyang Zhang et al.

Retrieval Augmented Generation (RAG) has been a powerful tool for Large Language Models (LLMs) to efficiently process overly lengthy contexts. However, recent LLMs like Gemini-1.5 and GPT-4 show exceptional capabilities to understand long contexts directly. We conduct a comprehensive comparison between RAG and long-context (LC) LLMs, aiming to leverage the strengths of both. We benchmark RAG and LC across various public datasets using three latest LLMs. Results reveal that when resourced sufficiently, LC consistently outperforms RAG in terms of average performance. However, RAG's significantly lower cost remains a distinct advantage. Based on this observation, we propose Self-Route, a simple yet effective method that routes queries to RAG or LC based on model self-reflection. Self-Route significantly reduces the computation cost while maintaining a comparable performance to LC. Our findings provide a guideline for long-context applications of LLMs using RAG and LC.

Haoran Li, Cheng Li, Weijian Huang et al.

Brain tumor segmentation based on multi-modal magnetic resonance imaging (MRI) plays a pivotal role in assisting brain cancer diagnosis, treatment, and postoperative evaluations. Despite the achieved inspiring performance by existing automatic segmentation methods, multi-modal MRI data are still unavailable in real-world clinical applications due to quite a few uncontrollable factors (e.g. different imaging protocols, data corruption, and patient condition limitations), which lead to a large performance drop during practical applications. In this work, we propose a Deeply supervIsed knowledGE tranSfer neTwork (DIGEST), which achieves accurate brain tumor segmentation under different modality-missing scenarios. Specifically, a knowledge transfer learning frame is constructed, enabling a student model to learn modality-shared semantic information from a teacher model pretrained with the complete multi-modal MRI data. To simulate all the possible modality-missing conditions under the given multi-modal data, we generate incomplete multi-modal MRI samples based on Bernoulli sampling. Finally, a deeply supervised knowledge transfer loss is designed to ensure the consistency of the teacher-student structure at different decoding stages, which helps the extraction of inherent and effective modality representations. Experiments on the BraTS 2020 dataset demonstrate that our method achieves promising results for the incomplete multi-modal MR image segmentation task.

Longfei Liu, Wen Guo, Shihua Huang et al.

Reducing false positives is essential for enhancing object detector performance, as reflected in the mean Average Precision (mAP) metric. Although object detectors have achieved notable improvements and high mAP scores on the COCO dataset, analysis reveals limited progress in addressing false positives caused by non-target visual clutter-background objects not included in the annotated categories. This issue is particularly critical in real-world applications, such as fire and smoke detection, where minimizing false alarms is crucial. In this study, we introduce COCO-FP, a new evaluation dataset derived from the ImageNet-1K dataset, designed to address this issue. By extending the original COCO validation dataset, COCO-FP specifically assesses object detectors' performance in mitigating background false positives. Our evaluation of both standard and advanced object detectors shows a significant number of false positives in both closed-set and open-set scenarios. For example, the AP50 metric for YOLOv9-E decreases from 72.8 to 65.7 when shifting from COCO to COCO-FP. The dataset is available at https://github.com/COCO-FP/COCO-FP.

Cheng Li, Mingyang Zhang, Qiaozhu Mei et al.

Facilitated by large language models (LLMs), personalized text generation has become a rapidly growing research direction. Most existing studies focus on designing specialized models for a particular domain, or they require fine-tuning the LLMs to generate personalized text. We consider a typical scenario in which the large language model, which generates personalized output, is frozen and can only be accessed through APIs. Under this constraint, all one can do is to improve the input text (i.e., text prompts) sent to the LLM, a procedure that is usually done manually. In this paper, we propose a novel method to automatically revise prompts for personalized text generation. The proposed method takes the initial prompts generated by a state-of-the-art, multistage framework for personalized generation and rewrites a few critical components that summarize and synthesize the personal context. The prompt rewriter employs a training paradigm that chains together supervised learning (SL) and reinforcement learning (RL), where SL reduces the search space of RL and RL facilitates end-to-end training of the rewriter. Using datasets from three representative domains, we demonstrate that the rewritten prompts outperform both the original prompts and the prompts optimized via supervised learning or reinforcement learning alone. In-depth analysis of the rewritten prompts shows that they are not only human readable, but also able to guide manual revision of prompts when there is limited resource to employ reinforcement learning to train the prompt rewriter, or when it is costly to deploy an automatic prompt rewriter for inference.

Yousuf Babiker M. Osman, Cheng Li, Weijian Huang et al.

Volumetric magnetic resonance (MR) image segmentation plays an important role in many clinical applications. Deep learning (DL) has recently achieved state-of-the-art or even human-level performance on various image segmentation tasks. Nevertheless, manually annotating volumetric MR images for DL model training is labor-exhaustive and time-consuming. In this work, we aim to train a semi-supervised and self-supervised collaborative learning framework for prostate 3D MR image segmentation while using extremely sparse annotations, for which the ground truth annotations are provided for just the central slice of each volumetric MR image. Specifically, semi-supervised learning and self-supervised learning methods are used to generate two independent sets of pseudo labels. These pseudo labels are then fused by Boolean operation to extract a more confident pseudo label set. The images with either manual or network self-generated labels are then employed to train a segmentation model for target volume extraction. Experimental results on a publicly available prostate MR image dataset demonstrate that, while requiring significantly less annotation effort, our framework generates very encouraging segmentation results. The proposed framework is very useful in clinical applications when training data with dense annotations are difficult to obtain.

Ruijin Liu, Ning Lu, Dapeng Chen et al.

We present PBFormer, an efficient yet powerful scene text detector that unifies the transformer with a novel text shape representation Polynomial Band (PB). The representation has four polynomial curves to fit a text's top, bottom, left, and right sides, which can capture a text with a complex shape by varying polynomial coefficients. PB has appealing features compared with conventional representations: 1) It can model different curvatures with a fixed number of parameters, while polygon-points-based methods need to utilize a different number of points. 2) It can distinguish adjacent or overlapping texts as they have apparent different curve coefficients, while segmentation-based or points-based methods suffer from adhesive spatial positions. PBFormer combines the PB with the transformer, which can directly generate smooth text contours sampled from predicted curves without interpolation. A parameter-free cross-scale pixel attention (CPA) module is employed to highlight the feature map of a suitable scale while suppressing the other feature maps. The simple operation can help detect small-scale texts and is compatible with the one-stage DETR framework, where no postprocessing exists for NMS. Furthermore, PBFormer is trained with a shape-contained loss, which not only enforces the piecewise alignment between the ground truth and the predicted curves but also makes curves' positions and shapes consistent with each other. Without bells and whistles about text pre-training, our method is superior to the previous state-of-the-art text detectors on the arbitrary-shaped text datasets.

Yifei Liu, Jicheng Wen, Yang Wang et al.

Scaling model size significantly challenges the deployment and inference of Large Language Models (LLMs). Due to the redundancy in LLM weights, recent research has focused on pushing weight-only quantization to extremely low-bit (even down to 2 bits). It reduces memory requirements, optimizes storage costs, and decreases memory bandwidth needs during inference. However, due to numerical representation limitations, traditional scalar-based weight quantization struggles to achieve such extreme low-bit. Recent research on Vector Quantization (VQ) for LLMs has demonstrated the potential for extremely low-bit model quantization by compressing vectors into indices using lookup tables. In this paper, we introduce Vector Post-Training Quantization (VPTQ) for extremely low-bit quantization of LLMs. We use Second-Order Optimization to formulate the LLM VQ problem and guide our quantization algorithm design by solving the optimization. We further refine the weights using Channel-Independent Second-Order Optimization for a granular VQ. In addition, by decomposing the optimization problem, we propose a brief and effective codebook initialization algorithm. We also extend VPTQ to support residual and outlier quantization, which enhances model accuracy and further compresses the model. Our experimental results show that VPTQ reduces model quantization perplexity by $0.01$-$0.34$ on LLaMA-2, $0.38$-$0.68$ on Mistral-7B, $4.41$-$7.34$ on LLaMA-3 over SOTA at 2-bit, with an average accuracy improvement of $0.79$-$1.5\%$ on LLaMA-2, $1\%$ on Mistral-7B, $11$-$22\%$ on LLaMA-3 on QA tasks on average. We only utilize $10.4$-$18.6\%$ of the quantization algorithm execution time, resulting in a $1.6$-$1.8\times$ increase in inference throughput compared to SOTA.

Zewen Jin, Congkun Ai, Guangpeng Zhang et al.

Modern Mixture-of-Experts (MoE) models increasingly rely on large-scale AI accelerator clusters for efficient training. Ascend NPUs expose heterogeneous on-chip compute resources, including matrix-oriented AIC units and vector-oriented AIV units with explicit cross-queue synchronization support. However, existing training frameworks largely execute MoE operators in a serialized kernel-by-kernel manner, leaving substantial heterogeneous parallelism underutilized. This paper presents HyperParallel-MoE, a compilation and scheduling framework for MoE training on Ascend NPUs. HyperParallel-MoE transforms operator-level MoE execution into a statically scheduled tile-level heterogeneous taskflow spanning AIC and AIV resources. It introduces AIV-driven one-sided communication to eliminate host-side collective synchronization, dependency-preserving tile task generation to unify communication and computation under a common task abstraction, and event-driven static scheduling to coordinate cross-queue execution with low runtime overhead. HyperParallel-MoE further executes the compiled taskflow within a unified runtime that concurrently drives AIC and AIV workers inside a single kernel launch, enabling fine-grained overlap among communication, matrix computation, and vector computation while preserving existing optimized operators. We implement HyperParallel-MoE in the MindSpore and MindFormers stack and evaluate it using DeepSeek-style MoE models on Ascend A3 clusters. Across multiple expert-parallel configurations, HyperParallel-MoE reduces Dispatch-to-Combine MoE-FFN latency by up to 1.58x, demonstrating that tile-level heterogeneous scheduling can substantially improve MoE training efficiency on modern NPUs.

Ping Gong, Yuxin Ma, Cheng Li et al.

In this paper, we primarily focus on understanding the data preprocessing pipeline for DNN Training in the public cloud. First, we run experiments to test the performance implications of the two major data preprocessing methods using either raw data or record files. The preliminary results show that data preprocessing is a clear bottleneck, even with the most efficient software and hardware configuration enabled by NVIDIA DALI, a high-optimized data preprocessing library. Second, we identify the potential causes, exercise a variety of optimization methods, and present their pros and cons. We hope this work will shed light on the new co-design of ``data storage, loading pipeline'' and ``training framework'' and flexible resource configurations between them so that the resources can be fully exploited and performance can be maximized.

Wenxin Fan, Jian Cheng, Cheng Li et al.

Deep learning has shown great potential in accelerating diffusion tensor imaging (DTI). Nevertheless, existing methods tend to suffer from Rician noise and eddy current, leading to detail loss in reconstructing the DTI-derived parametric maps especially when sparsely sampled q-space data are used. To address this, this paper proposes a novel method, AID-DTI (\textbf{A}ccelerating h\textbf{I}gh fi\textbf{D}elity \textbf{D}iffusion \textbf{T}ensor \textbf{I}maging), to facilitate fast and accurate DTI with only six measurements. AID-DTI is equipped with a newly designed Singular Value Decomposition-based regularizer, which can effectively capture fine details while suppressing noise during network training by exploiting the correlation across DTI-derived parameters. Additionally, we introduce a Nesterov-based adaptive learning algorithm that optimizes the regularization parameter dynamically to enhance the performance. AID-DTI is an extendable framework capable of incorporating flexible network architecture. Experimental results on Human Connectome Project (HCP) data consistently demonstrate that the proposed method estimates DTI parameter maps with fine-grained details and outperforms other state-of-the-art methods both quantitatively and qualitatively.

Xue Liu, Juan Zou, Xiawu Zheng et al.

Magnetic Resonance Imaging (MRI) has become an important technique in the clinic for the visualization, detection, and diagnosis of various diseases. However, one bottleneck limitation of MRI is the relatively slow data acquisition process. Fast MRI based on k-space undersampling and high-quality image reconstruction has been widely utilized, and many deep learning-based methods have been developed in recent years. Although promising results have been achieved, most existing methods require fully-sampled reference data for training the deep learning models. Unfortunately, fully-sampled MRI data are difficult if not impossible to obtain in real-world applications. To address this issue, we propose a data refinement framework for self-supervised MR image reconstruction. Specifically, we first analyze the reason of the performance gap between self-supervised and supervised methods and identify that the bias in the training datasets between the two is one major factor. Then, we design an effective self-supervised training data refinement method to reduce this data bias. With the data refinement, an enhanced self-supervised MR image reconstruction framework is developed to prompt accurate MR imaging. We evaluate our method on an in-vivo MRI dataset. Experimental results show that without utilizing any fully sampled MRI data, our self-supervised framework possesses strong capabilities in capturing image details and structures at high acceleration factors.

Cheng Li, Yousuf Babiker M. Osman, Weijian Huang et al.

Multi-parametric magnetic resonance (MR) imaging is an indispensable tool in the clinic. Consequently, automatic volume-of-interest segmentation based on multi-parametric MR imaging is crucial for computer-aided disease diagnosis, treatment planning, and prognosis monitoring. Despite the extensive studies conducted in deep learning-based medical image analysis, further investigations are still required to effectively exploit the information provided by different imaging parameters. How to fuse the information is a key question in this field. Here, we propose an uncertainty-aware multi-parametric MR image feature fusion method to fully exploit the information for enhanced 3D image segmentation. Uncertainties in the independent predictions of individual modalities are utilized to guide the fusion of multi-modal image features. Extensive experiments on two datasets, one for brain tissue segmentation and the other for abdominal multi-organ segmentation, have been conducted, and our proposed method achieves better segmentation performance when compared to existing models.

Yeqi Wang, Weijian Huang, Cheng Li et al.

Multi-contrast magnetic resonance imaging (MRI)-based automatic auxiliary glioma diagnosis plays an important role in the clinic. Contrast-enhanced MRI sequences (e.g., contrast-enhanced T1-weighted imaging) were utilized in most of the existing relevant studies, in which remarkable diagnosis results have been reported. Nevertheless, acquiring contrast-enhanced MRI data is sometimes not feasible due to the patients physiological limitations. Furthermore, it is more time-consuming and costly to collect contrast-enhanced MRI data in the clinic. In this paper, we propose an adaptive PromptNet to address these issues. Specifically, a PromptNet for glioma grading utilizing only non-enhanced MRI data has been constructed. PromptNet receives constraints from features of contrast-enhanced MR data during training through a designed prompt loss. To further boost the performance, an adaptive strategy is designed to dynamically weight the prompt loss in a sample-based manner. As a result, PromptNet is capable of dealing with more difficult samples. The effectiveness of our method is evaluated on a widely-used BraTS2020 dataset, and competitive glioma grading performance on NE-MRI data is achieved.

Yaqing Wang, Jiepu Jiang, Mingyang Zhang et al.

Personalized text generation presents a specialized mechanism for delivering content that is specific to a user's personal context. While the research progress in this area has been rapid, evaluation still presents a challenge. Traditional automated metrics such as BLEU and ROUGE primarily measure lexical similarity to human-written references, and are not able to distinguish personalization from other subtle semantic aspects, thus falling short of capturing the nuances of personalized generated content quality. On the other hand, human judgments are costly to obtain, especially in the realm of personalized evaluation. Inspired by these challenges, we explore the use of large language models (LLMs) for evaluating personalized text generation, and examine their ability to understand nuanced user context. We present AuPEL, a novel evaluation method that distills three major semantic aspects of the generated text: personalization, quality and relevance, and automatically measures these aspects. To validate the effectiveness of AuPEL, we design carefully controlled experiments and compare the accuracy of the evaluation judgments made by LLMs versus that of judgements made by human annotators, and conduct rigorous analyses of the consistency and sensitivity of the proposed metric. We find that, compared to existing evaluation metrics, AuPEL not only distinguishes and ranks models based on their personalization abilities more accurately, but also presents commendable consistency and efficiency for this task. Our work suggests that using LLMs as the evaluators of personalized text generation is superior to traditional text similarity metrics, even though interesting new challenges still remain.

Cheng Li, Mengzhou Chen, Jindong Wang et al.

Large language models (LLMs) are reported to be partial to certain cultures owing to the training data dominance from the English corpora. Since multilingual cultural data are often expensive to collect, existing efforts handle this by prompt engineering or culture-specific pre-training. However, they might overlook the knowledge deficiency of low-resource culture and require extensive computing resources. In this paper, we propose CultureLLM, a cost-effective solution to incorporate cultural differences into LLMs. CultureLLM adopts World Value Survey (WVS) as seed data and generates semantically equivalent training data via the proposed semantic data augmentation. Using only 50 seed samples from WVS with augmented data, we fine-tune culture-specific LLMs and one unified model (CultureLLM-One) for 9 cultures covering rich and low-resource languages. Extensive experiments on 60 culture-related datasets demonstrate that CultureLLM significantly outperforms various counterparts such as GPT-3.5 (by 8.1%) and Gemini Pro (by 9.5%) with comparable performance to GPT-4 or even better. Our human study shows that the generated samples are semantically equivalent to the original samples, providing an effective solution for LLMs augmentation. Code is released at https://github.com/Scarelette/CultureLLM.

Kimi Team, Yifan Bai, Yiping Bao et al. · tsinghua

We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated parameters and 1 trillion total parameters. We propose the MuonClip optimizer, which improves upon Muon with a novel QK-clip technique to address training instability while enjoying the advanced token efficiency of Muon. Based on MuonClip, K2 was pre-trained on 15.5 trillion tokens with zero loss spike. During post-training, K2 undergoes a multi-stage post-training process, highlighted by a large-scale agentic data synthesis pipeline and a joint reinforcement learning (RL) stage, where the model improves its capabilities through interactions with real and synthetic environments. Kimi K2 achieves state-of-the-art performance among open-source non-thinking models, with strengths in agentic capabilities. Notably, K2 obtains 66.1 on Tau2-Bench, 76.5 on ACEBench (En), 65.8 on SWE-Bench Verified, and 47.3 on SWE-Bench Multilingual -- surpassing most open and closed-sourced baselines in non-thinking settings. It also exhibits strong capabilities in coding, mathematics, and reasoning tasks, with a score of 53.7 on LiveCodeBench v6, 49.5 on AIME 2025, 75.1 on GPQA-Diamond, and 27.1 on OJBench, all without extended thinking. These results position Kimi K2 as one of the most capable open-source large language models to date, particularly in software engineering and agentic tasks. We release our base and post-trained model checkpoints to facilitate future research and applications of agentic intelligence.

Baomin Wang, Geng Hu, Dan Chen et al.

Accurately and quickly binuclear cell (BC) detection plays a significant role in predicting the risk of leukemia and other malignant tumors. However, manual microscopy counting is time-consuming and lacks objectivity. Moreover, with the limitation of staining quality and diversity of morphology features in BC microscopy whole slide images (WSIs), traditional image processing approaches are helpless. To overcome this challenge, we propose a two-stage detection method inspired by the structure prior of BC based on deep learning, which cascades to implement BCs coarse detection at the WSI-level and fine-grained classification in patch-level. The coarse detection network is a multi-task detection framework based on circular bounding boxes for cells detection, and central key points for nucleus detection. The circle representation reduces the degrees of freedom, mitigates the effect of surrounding impurities compared to usual rectangular boxes and can be rotation invariant in WSI. Detecting key points in the nucleus can assist network perception and be used for unsupervised color layer segmentation in later fine-grained classification. The fine classification network consists of a background region suppression module based on color layer mask supervision and a key region selection module based on a transformer due to its global modeling capability. Additionally, an unsupervised and unpaired cytoplasm generator network is firstly proposed to expand the long-tailed distribution dataset. Finally, experiments are performed on BC multicenter datasets. The proposed BC fine detection method outperforms other benchmarks in almost all the evaluation criteria, providing clarification and support for tasks such as cancer screenings.

Seed, Baisheng Li, Banggu Wu et al.

We introduce Virtual Width Networks (VWN), a framework that delivers the benefits of wider representations without incurring the quadratic cost of increasing the hidden size. VWN decouples representational width from backbone width, expanding the embedding space while keeping backbone compute nearly constant. In our large-scale experiment, an 8-times expansion accelerates optimization by over 2 times for next-token and 3 times for next-2-token prediction. The advantage amplifies over training as both the loss gap grows and the convergence-speedup ratio increases, showing that VWN is not only token-efficient but also increasingly effective with scale. Moreover, we identify an approximately log-linear scaling relation between virtual width and loss reduction, offering an initial empirical basis and motivation for exploring virtual-width scaling as a new dimension of large-model efficiency.

Xu Wang, Cheng Li, Yi Chang et al.

Large Language Models (LLMs) have become integral to a wide spectrum of applications, ranging from traditional computing tasks to advanced artificial intelligence (AI) applications. This widespread adoption has spurred extensive research into LLMs across various disciplines, including the social sciences. Notably, studies have revealed that LLMs possess emotional intelligence, which can be further developed through positive emotional stimuli. This discovery raises an intriguing question: can negative emotions similarly influence LLMs, potentially enhancing their performance? In response to this question, we introduce NegativePrompt, a novel approach underpinned by psychological principles, involving ten specifically designed negative emotional stimuli. We embark on rigorous experimental evaluations of five LLMs including Flan-T5-Large, Vicuna, Llama 2, ChatGPT, and GPT-4, across a set of 45 tasks. The results are revealing: NegativePrompt markedly enhances the performance of LLMs, evidenced by relative improvements of 12.89% in Instruction Induction tasks and 46.25% in BIG-Bench tasks. Moreover, we conduct attention visualization experiments to decipher the underlying mechanisms of NegativePrompt's influence. Our research contributes significantly to the understanding of LLMs and emotion interaction, demonstrating the practical efficacy of NegativePrompt as an emotion-driven method and offering novel insights for the enhancement of LLMs in real-world applications. The code is available at https://github.com/wangxu0820/NegativePrompt.

Cheng Li, Weijian Huang, Jiarun Liu et al.

The integration of medical imaging and clinical text has enabled the emergence of generalist artificial intelligence (AI) systems for healthcare. However, pervasive biases, such as imbalanced disease prevalence, skewed anatomical region distributions, heterogeneous imaging protocols, and demographic disparities, pose significant challenges to the fairness and reliability of vision-language systems in real-world clinical settings. Here we present BiasCareVL, a bias-aware multimodal learning framework that introduces bias control directly into model design, rather than treating it as a post hoc correction. BiasCareVL incorporates adaptive uncertainty modeling with optional human-in-the-loop refinement to regulate the influence of dominant data patterns and to promote equitable reasoning under distributional imbalance. Trained on 3.44 million samples spanning over 15 imaging modalities, the framework supports diverse clinical tasks, including visual question answering, disease classification, segmentation, and report generation within a unified representation space. Across eight public benchmarks covering dermatology, oncology, radiology, and pathology, BiasCareVL consistently outperforms 20 state-of-the-art methods, with pronounced gains in clinically challenging scenarios, including over 10% accuracy improvement in multi-class skin lesion diagnosis and more than 20% Dice improvement in small tumor segmentation. Furthermore, BiasCareVL achieves diagnostic performance exceeding human accuracy with substantially reduced time requirements when evaluated with board-certified radiologists. By open-sourcing BiasCareVL, we aim to promote a transparent, reproducible, and equitable future for AI in healthcare, paving the way for general-purpose, trustworthy, and clinically reliable AI systems.