Peisong Wang

Yuhan Li, Zhixun Li, Peisong Wang et al.

Graph plays a significant role in representing and analyzing complex relationships in real-world applications such as citation networks, social networks, and biological data. Recently, Large Language Models (LLMs), which have achieved tremendous success in various domains, have also been leveraged in graph-related tasks to surpass traditional Graph Neural Networks (GNNs) based methods and yield state-of-the-art performance. In this survey, we first present a comprehensive review and analysis of existing methods that integrate LLMs with graphs. First of all, we propose a new taxonomy, which organizes existing methods into three categories based on the role (i.e., enhancer, predictor, and alignment component) played by LLMs in graph-related tasks. Then we systematically survey the representative methods along the three categories of the taxonomy. Finally, we discuss the remaining limitations of existing studies and highlight promising avenues for future research. The relevant papers are summarized and will be consistently updated at: https://github.com/yhLeeee/Awesome-LLMs-in-Graph-tasks.

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.

Yuhan Li, Peisong Wang, Xiao Zhu et al.

The emergence of large language models (LLMs) has revolutionized the way we interact with graphs, leading to a new paradigm called GraphLLM. Despite the rapid development of GraphLLM methods in recent years, the progress and understanding of this field remain unclear due to the lack of a benchmark with consistent experimental protocols. To bridge this gap, we introduce GLBench, the first comprehensive benchmark for evaluating GraphLLM methods in both supervised and zero-shot scenarios. GLBench provides a fair and thorough evaluation of different categories of GraphLLM methods, along with traditional baselines such as graph neural networks. Through extensive experiments on a collection of real-world datasets with consistent data processing and splitting strategies, we have uncovered several key findings. Firstly, GraphLLM methods outperform traditional baselines in supervised settings, with LLM-as-enhancers showing the most robust performance. However, using LLMs as predictors is less effective and often leads to uncontrollable output issues. We also notice that no clear scaling laws exist for current GraphLLM methods. In addition, both structures and semantics are crucial for effective zero-shot transfer, and our proposed simple baseline can even outperform several models tailored for zero-shot scenarios. The data and code of the benchmark can be found at https://github.com/NineAbyss/GLBench.

Weixiang Xu, Xiangyu He, Tianli Zhao et al.

Large neural networks are difficult to deploy on mobile devices because of intensive computation and storage. To alleviate it, we study ternarization, a balance between efficiency and accuracy that quantizes both weights and activations into ternary values. In previous ternarized neural networks, a hard threshold Δ is introduced to determine quantization intervals. Although the selection of Δ greatly affects the training results, previous works estimate Δ via an approximation or treat it as a hyper-parameter, which is suboptimal. In this paper, we present the Soft Threshold Ternary Networks (STTN), which enables the model to automatically determine quantization intervals instead of depending on a hard threshold. Concretely, we replace the original ternary kernel with the addition of two binary kernels at training time, where ternary values are determined by the combination of two corresponding binary values. At inference time, we add up the two binary kernels to obtain a single ternary kernel. Our method dramatically outperforms current state-of-the-arts, lowering the performance gap between full-precision networks and extreme low bit networks. Experiments on ImageNet with ResNet-18 (Top-1 66.2%) achieves new state-of-the-art. Update: In this version, we further fine-tune the experimental hyperparameters and training procedure. The latest STTN shows that ResNet-18 with ternary weights and ternary activations achieves up to 68.2% Top-1 accuracy on ImageNet. Code is available at: github.com/WeixiangXu/STTN.

Changwang Mei, Peisong Wang, Zekun Li et al.

Visual Autoregressive (VAR) models deliver high-quality image generation but suffer from significant inference latency at high resolutions. Recent acceleration approaches most rely on heuristic measures with layer features to prune tokens. Such heuristics are sensitive to complex contextual semantics, leading to inaccurate identification of redundant computation and poor adaptability across prompts. We rethink redundancy in VAR from the perspective of its impact on pixel-space generation and introduce Latent Discrepancy. This unified metric quantifies a token's contribution by measuring the change in model states during generation. Our analysis shows that redundancy is more accurately identified when guided by image latent or pixel-space signals. We further observed that in classifier-free guidance (CFG), the convergence trend of the discrepancy between conditional and unconditional branches exhibits high dynamics with different prompts. Based on these findings, we propose LD-Pruning (Latent Discrepancy Pruning), a training-free framework that removes redundancy via latent discrepancy by integrating decoding-free region selection and adaptive unconditional-branch skipping. Extensive experiments show that LD-Pruning substantially reduces inference latency while maintaining high generation quality, achieving up to 2.35x speedup on Infinity-8B.

Zeyu Zhu, Peisong Wang, Qinghao Hu et al.

Graph Neural Networks (GNNs) have shown great superiority on non-Euclidean graph data, achieving ground-breaking performance on various graph-related tasks. As a practical solution to train GNN on large graphs with billions of nodes and edges, the sampling-based training is widely adopted by existing training frameworks. However, through an in-depth analysis, we observe that the efficiency of existing sampling-based training frameworks is still limited due to the key bottlenecks lying in all three phases of sampling-based training, i.e., subgraph sample, memory IO, and computation. To this end, we propose FastGL, a GPU-efficient Framework for accelerating sampling-based training of GNN at Large scale by simultaneously optimizing all above three phases, taking into account both GPU characteristics and graph structure. Specifically, by exploiting the inherent overlap within graph structures, FastGL develops the Match-Reorder strategy to reduce the data traffic, which accelerates the memory IO without incurring any GPU memory overhead. Additionally, FastGL leverages a Memory-Aware computation method, harnessing the GPU memory's hierarchical nature to mitigate irregular data access during computation. FastGL further incorporates the Fused-Map approach aimed at diminishing the synchronization overhead during sampling. Extensive experiments demonstrate that FastGL can achieve an average speedup of 11.8x, 2.2x and 1.5x over the state-of-the-art frameworks PyG, DGL, and GNNLab, respectively.Our code is available at https://github.com/a1bc2def6g/fastgl-ae.

Zekun Li, Ning Wang, Tongxin Bai et al.

Visual AutoRegressive (VAR) modeling has garnered significant attention for its innovative next-scale prediction paradigm. However, mainstream VAR paradigms attend to all tokens across historical scales at each autoregressive step. As the next scale resolution grows, the computational complexity of attention increases quartically with resolution, causing substantial latency. Prior accelerations often skip high-resolution scales, which speeds up inference but discards high-frequency details and harms image quality. To address these problems, we present SparVAR, a training-free acceleration framework that exploits three properties of VAR attention: (i) strong attention sinks, (ii) cross-scale activation similarity, and (iii) pronounced locality. Specifically, we dynamically predict the sparse attention pattern of later high-resolution scales from a sparse decision scale, and construct scale self-similar sparse attention via an efficient index-mapping mechanism, enabling high-efficiency sparse attention computation at large scales. Furthermore, we propose cross-scale local sparse attention and implement an efficient block-wise sparse kernel, which achieves $\mathbf{> 5\times}$ faster forward speed than FlashAttention. Extensive experiments demonstrate that the proposed SparseVAR can reduce the generation time of an 8B model producing $1024\times1024$ high-resolution images to the 1s, without skipping the last scales. Compared with the VAR baseline accelerated by FlashAttention, our method achieves a $\mathbf{1.57\times}$ speed-up while preserving almost all high-frequency details. When combined with existing scale-skipping strategies, SparseVAR attains up to a $\mathbf{2.28\times}$ acceleration, while maintaining competitive visual generation quality. Code is available at https://github.com/CAS-CLab/SparVAR.

Yuantian Shao, Yuanteng Chen, Peisong Wang et al.

Quantization plays a crucial role in accelerating the inference of large-scale models, and rotational matrices have been shown to effectively improve quantization performance by smoothing outliers. However, end-to-end fine-tuning of rotational optimization algorithms incurs high computational costs and is prone to overfitting. To address this challenge, we propose an efficient distribution-aware rotational calibration method, DartQuant, which reduces the complexity of rotational optimization by constraining the distribution of the activations after rotation. This approach also effectively reduces reliance on task-specific losses, thereby mitigating the risk of overfitting. Additionally, we introduce the QR-Orth optimization scheme, which replaces expensive alternating optimization with a more efficient solution. In a variety of model quantization experiments, DartQuant demonstrates superior performance. Compared to existing methods, it achieves 47$\times$ acceleration and 10$\times$ memory savings for rotational optimization on a 70B model. Furthermore, it is the first to successfully complete rotational calibration for a 70B model on a single 3090 GPU, making quantization of large language models feasible in resource-constrained environments. Code is available at https://github.com/CAS-CLab/DartQuant.git.

Anda Cheng, Peisong Wang, Xi Sheryl Zhang et al.

User-level differential privacy (DP) provides certifiable privacy guarantees to the information that is specific to any user's data in federated learning. Existing methods that ensure user-level DP come at the cost of severe accuracy decrease. In this paper, we study the cause of model performance degradation in federated learning under user-level DP guarantee. We find the key to solving this issue is to naturally restrict the norm of local updates before executing operations that guarantee DP. To this end, we propose two techniques, Bounded Local Update Regularization and Local Update Sparsification, to increase model quality without sacrificing privacy. We provide theoretical analysis on the convergence of our framework and give rigorous privacy guarantees. Extensive experiments show that our framework significantly improves the privacy-utility trade-off over the state-of-the-arts for federated learning with user-level DP guarantee.

Zhongjian Qiao, Jiafei Lyu, Chenjia Bai et al.

Cross-domain offline reinforcement learning (RL) aims to learn a policy in the target domain with a limited target domain dataset and a source domain dataset that exhibits a dynamics shift. Training directly on the original source dataset typically leads to performance collapse. Recent studies perform data filtering from the perspective of dynamics alignment or value alignment to enable efficient policy transfer. However, these studies are typically validated on single-domain or single-behavior-policy source datasets. In this work, we explore a more general heterogeneous cross-domain offline RL setting, where the source datasets may be collected from multiple source domains by diverse behavior policies. We first uncover a critical yet overlooked issue in this setting: value misassignment. Empirically and theoretically, we demonstrate that value misassignment can undermine value alignment, mislead data filtering toward selecting suboptimal samples, and loosen the suboptimality gap, thereby degrading the agent's performance. To address this issue, we propose V2A, which integrates dynamics alignment, value alignment, and value assignment. V2A first employs temporally-consistent modality representation learning to extract dynamics modalities from the source dataset, followed by modality-aware advantage learning to rectify value alignment. Finally, it adopts a data filtering paradigm to selectively share source data for policy learning. Empirical results show that V2A significantly outperforms strong baseline methods under general heterogeneous cross-domain offline RL settings.

Yuhan Li, Peisong Wang, Zhixun Li et al.

With the development of foundation models such as large language models, zero-shot transfer learning has become increasingly significant. This is highlighted by the generative capabilities of NLP models like GPT-4, and the retrieval-based approaches of CV models like CLIP, both of which effectively bridge the gap between seen and unseen data. In the realm of graph learning, the continuous emergence of new graphs and the challenges of human labeling also amplify the necessity for zero-shot transfer learning, driving the exploration of approaches that can generalize across diverse graph data without necessitating dataset-specific and label-specific fine-tuning. In this study, we extend such paradigms to zero-shot transferability in graphs by introducing ZeroG, a new framework tailored to enable cross-dataset generalization. Addressing the inherent challenges such as feature misalignment, mismatched label spaces, and negative transfer, we leverage a language model to encode both node attributes and class semantics, ensuring consistent feature dimensions across datasets. We also propose a prompt-based subgraph sampling module that enriches the semantic information and structure information of extracted subgraphs using prompting nodes and neighborhood aggregation, respectively. We further adopt a lightweight fine-tuning strategy that reduces the risk of overfitting and maintains the zero-shot learning efficacy of the language model. The results underscore the effectiveness of our model in achieving significant cross-dataset zero-shot transferability, opening pathways for the development of graph foundation models. Codes and data are available at https://github.com/NineAbyss/ZeroG.

Peisong Wang, Bowen Liu, Zehua Li et al.

Large language models still struggle with contest-level programming, while many agentic remedies rely on massive inference-time sampling or expensive multi-stage post-training. We study when execution feedback reliably helps an LLM CP solver and which mechanisms govern the gains. We model feedback-driven solving as a calibrated stopped process and identify three quantities: false-admission risk, program-level evidence against bad programs, and the active-state success hazard. Under held-out trace calibration and selection from a pre-declared finite controller manifest, the resulting structural certificate lower-bounds the clean success probability before false admission. We instantiate mechanisms targeting these quantities as Dual-Granularity Verification, Test Augmentation, and Experience-Driven Self-Evolving, yielding CP-Agent. Without updating any parameters, CP-Agent raises Pass@1 from 25.8\% to 48.5\% on LiveCodeBench Pro and improves Refine@5 by 11.0\% on ICPC-Eval. Across three LLM backbones, CP-Agent lies on the cost--accuracy efficiency frontier, and ablations show that each component primarily affects its corresponding certificate quantity.

Yingying Deng, Xiangyu He, Changwang Mei et al.

Though Rectified Flows (ReFlows) with distillation offers a promising way for fast sampling, its fast inversion transforms images back to structured noise for recovery and following editing remains unsolved. This paper introduces FireFlow, a simple yet effective zero-shot approach that inherits the startling capacity of ReFlow-based models (such as FLUX) in generation while extending its capabilities to accurate inversion and editing in $8$ steps. We first demonstrate that a carefully designed numerical solver is pivotal for ReFlow inversion, enabling accurate inversion and reconstruction with the precision of a second-order solver while maintaining the practical efficiency of a first-order Euler method. This solver achieves a $3\times$ runtime speedup compared to state-of-the-art ReFlow inversion and editing techniques, while delivering smaller reconstruction errors and superior editing results in a training-free mode. The code is available at $\href{https://github.com/HolmesShuan/FireFlow}{this URL}$.

Yuanteng Chen, Peisong Wang, Nanxin Zeng et al.

Test-time scaling improves LLM performance by generating multiple candidate solutions, yet token-level sampling requires temperature tuning that trades off diversity against stability. Fine-grained MoE, featuring hundreds of well-trained experts per layer and multi-expert activation per token, offers an unexplored alternative through its rich routing space. We empirically characterize fine-grained MoE routing and uncover an informative pattern: router scores exhibit a certain head of high-confidence experts followed by an uncertain tail of low-confidence candidates. While single-run greedy accuracy remains stable when fewer experts are activated, multi-sample pass@n degrades significantly-suggesting that the certain head governs core reasoning capability while the uncertain tail correlates with reasoning diversity. Motivated by these findings, we propose Expert-Sample, a training-free method that preserves high-confidence selections while injecting controlled stochasticity into the uncertain tail, enabling diverse generation without destabilizing outputs. Evaluated on multiple fine-grained MoE models across math, knowledge reasoning, and code tasks, Expert-Sample consistently improves pass@n and verification-based accuracy. On Qwen3-30B-A3B-Instruct evaluated on GPQA-Diamond with 32 parallel samples, pass@32 rises from 85.4% to 91.9%, and accuracy improves from 59.1% to 62.6% with Best-of-N verification.

Zeyu Zhu, Gang Li, Peisong Wang et al.

Mixture of Experts (MoE) architectures significantly enhance the capacity of LLMs without proportional increases in computation, but at the cost of a vast parameter size. Offloading MoE expert parameters to host memory and leveraging both CPU and GPU computation has recently emerged as a promising direction to support such models on resourceconstrained local PC platforms. While promising, we notice that existing approaches mismatch the dynamic nature of expert workloads, which leads to three fundamental inefficiencies: (1) Static expert assignment causes severe CPUGPU load imbalance, underutilizing CPU and GPU resources; (2) Existing prefetching techniques fail to accurately predict high-workload experts, leading to costly inaccurate prefetches; (3) GPU cache policies neglect workload dynamics, resulting in poor hit rates and limited effectiveness. To address these challenges, we propose DALI, a workloaDAware offLoadIng framework for efficient MoE inference on local PCs. To fully utilize hardware resources, DALI first dynamically assigns experts to CPU or GPU by modeling assignment as a 0-1 integer optimization problem and solving it efficiently using a Greedy Assignment strategy at runtime. To improve prefetching accuracy, we develop a Residual-Based Prefetching method leveraging inter-layer residual information to accurately predict high-workload experts. Additionally, we introduce a Workload-Aware Cache Replacement policy that exploits temporal correlation in expert activations to improve GPU cache efficiency. By evaluating across various MoE models and settings, DALI achieves significant speedups in the both prefill and decoding phases over the state-of-the-art offloading frameworks.

Meng Li, Peisong Wang, Yuantian Shao et al.

Large Language Models (LLMs) achieve strong performance across diverse tasks but face deployment challenges due to their massive size. Structured pruning offers acceleration benefits but leads to significant performance degradation. Recent PCA-based pruning methods have alleviated this issue by retaining key activation components, but are only applied between modules in order to fuse the transformation matrix, which introduces extra parameters and severely disrupts activation distributions due to residual connections. To address these issues, we propose IntraSlice, a framework that applies block-wise module-intra PCA compression pruning. By leveraging the structural characteristics of Transformer modules, we design an approximate PCA method whose transformation matrices can be fully fused into the model without additional parameters. We also introduce a PCA-based global pruning ratio estimator that further considers the distribution of compressed activations, building on conventional module importance. We validate our method on Llama2, Llama3, and Phi series across various language benchmarks. Experimental results demonstrate that our approach achieves superior compression performance compared to recent baselines at the same compression ratio or inference speed.

Yuantian Shao, Peisong Wang, Yuanteng Chen et al.

Large language models (LLMs) have achieved remarkable success, but their rapidly growing scale imposes prohibitive costs in memory, computation, and energy. Post-training quantization (PTQ) is a promising solution for efficient deployment, yet achieving accurate W4A4 quantization remains an open challenge. While most existing methods are designed for INT4 formats, the emergence of MXFP4 -- a new FP4 format with various hardware support (NVIDIA, AMD, Intel)-- raises questions about the applicability of current techniques. In this work, we establish a comprehensive benchmark of PTQ methods under the MXFP4 format. Through systematic evaluation, we find that methods like GPTQ consistently deliver strong performance, whereas rotation-based approaches, which are almost used by all state-of-the-art approaches, suffer from severe incompatibility with MXFP4. We further provide the first in-depth analysis of this conflict, tracing its root to a fundamental mismatch between MXFP4's PoT (power-of-two) block scaling and the redistribution of outlier energy via global rotation. Building on this insight, we propose a simple yet effective block rotation strategy that adapts rotation-based methods to MXFP4, leading to substantial accuracy improvements across diverse LLMs. Our findings not only offer clear guidance for practitioners but also set a foundation for advancing PTQ research under emerging low-precision formats.

Ruotian Ma, Peisong Wang, Cheng Liu et al.

Recent studies have demonstrated the effectiveness of LLM test-time scaling. However, existing approaches to incentivize LLMs' deep thinking abilities generally require large-scale data or significant training efforts. Meanwhile, it remains unclear how to improve the thinking abilities of less powerful base models. In this work, we introduce S$^2$R, an efficient framework that enhances LLM reasoning by teaching models to self-verify and self-correct during inference. Specifically, we first initialize LLMs with iterative self-verification and self-correction behaviors through supervised fine-tuning on carefully curated data. The self-verification and self-correction skills are then further strengthened by both outcome-level and process-level reinforcement learning, with minimized resource requirements, enabling the model to adaptively refine its reasoning process during inference. Our results demonstrate that, with only 3.1k self-verifying and self-correcting behavior initialization samples, Qwen2.5-math-7B achieves an accuracy improvement from 51.0\% to 81.6\%, outperforming models trained on an equivalent amount of long-CoT distilled data. Extensive experiments and analysis based on three base models across both in-domain and out-of-domain benchmarks validate the effectiveness of S$^2$R. Our code and data are available at https://github.com/NineAbyss/S2R.

Bang Zhang, Ruotian Ma, Qingxuan Jiang et al.

Assessing how well a large language model (LLM) understands human, rather than merely text, remains an open challenge. To bridge the gap, we introduce Sentient Agent as a Judge (SAGE), an automated evaluation framework that measures an LLM's higher-order social cognition. SAGE instantiates a Sentient Agent that simulates human-like emotional changes and inner thoughts during interaction, providing a more realistic evaluation of the tested model in multi-turn conversations. At every turn, the agent reasons about (i) how its emotion changes, (ii) how it feels, and (iii) how it should reply, yielding a numerical emotion trajectory and interpretable inner thoughts. Experiments on 100 supportive-dialogue scenarios show that the final Sentient emotion score correlates strongly with Barrett-Lennard Relationship Inventory (BLRI) ratings and utterance-level empathy metrics, validating psychological fidelity. We also build a public Sentient Leaderboard covering 18 commercial and open-source models that uncovers substantial gaps (up to 4x) between frontier systems (GPT-4o-Latest, Gemini2.5-Pro) and earlier baselines, gaps not reflected in conventional leaderboards (e.g., Arena). SAGE thus provides a principled, scalable and interpretable tool for tracking progress toward genuinely empathetic and socially adept language agents.

Xingting Yao, Qinghao Hu, Fei Zhou et al.

Multi-bit spiking neural networks (SNNs) have recently become a heated research spot, pursuing energy-efficient and high-accurate AI. However, with more bits involved, the associated memory and computation demands escalate to the point where the performance improvements become disproportionate. Based on the insight that different layers demonstrate different importance and extra bits could be wasted and interfering, this paper presents an adaptive bit allocation strategy for direct-trained SNNs, achieving fine-grained layer-wise allocation of memory and computation resources. Thus, SNN's efficiency and accuracy can be improved. Specifically, we parametrize the temporal lengths and the bit widths of weights and spikes, and make them learnable and controllable through gradients. To address the challenges caused by changeable bit widths and temporal lengths, we propose the refined spiking neuron, which can handle different temporal lengths, enable the derivation of gradients for temporal lengths, and suit spike quantization better. In addition, we theoretically formulate the step-size mismatch problem of learnable bit widths, which may incur severe quantization errors to SNN, and accordingly propose the step-size renewal mechanism to alleviate this issue. Experiments on various datasets, including the static CIFAR and ImageNet datasets and the dynamic CIFAR-DVS, DVS-GESTURE, and SHD datasets, demonstrate that our methods can reduce the overall memory and computation cost while achieving higher accuracy. Particularly, our SEWResNet-34 can achieve a 2.69% accuracy gain and 4.16x lower bit budgets over the advanced baseline work on ImageNet. This work will be open-sourced.

Tianqi Chen, Zhe Li, Weixiang Xu et al.

Large language models (LLMs) have achieved remarkable performance on Natural Language Processing (NLP) tasks, but they are hindered by high computational costs and memory requirements. Ternarization, an extreme form of quantization, offers a solution by reducing memory usage and enabling energy-efficient floating-point additions. However, applying ternarization to LLMs faces challenges stemming from outliers in both weights and activations. In this work, observing asymmetric outliers and non-zero means in weights, we introduce Dual Learnable Ternarization (DLT), which enables both scales and shifts to be learnable. We also propose Outlier-Friendly Feature Knowledge Distillation (OFF) to recover the information lost in extremely low-bit quantization. The proposed OFF can incorporate semantic information and is insensitive to outliers. At the core of OFF is maximizing the mutual information between features in ternarized and floating-point models using cosine similarity. Extensive experiments demonstrate that our TernaryLLM surpasses previous low-bit quantization methods on the standard text generation and zero-shot benchmarks for different LLM families. Specifically, for one of the most powerful open-source models, LLaMA-3, our approach (W1.58A16) outperforms the previous state-of-the-art method (W2A16) by 5.8 in terms of perplexity on C4 and by 8.2% in terms of average accuracy on zero-shot tasks.

Xiangyu He, Zitao Mo, Qiang Chen et al.

Many successful learning targets such as minimizing dice loss and cross-entropy loss have enabled unprecedented breakthroughs in segmentation tasks. Beyond these semantic metrics, this paper aims to introduce location supervision into semantic segmentation. Based on this idea, we present a Location-aware Upsampling (LaU) that adaptively refines the interpolating coordinates with trainable offsets. Then, location-aware losses are established by encouraging pixels to move towards well-classified locations. An LaU is offset prediction coupled with interpolation, which is trained end-to-end to generate confidence score at each position from coarse to fine. Guided by location-aware losses, the new module can replace its plain counterpart (\textit{e.g.}, bilinear upsampling) in a plug-and-play manner to further boost the leading encoder-decoder approaches. Extensive experiments validate the consistent improvement over the state-of-the-art methods on benchmark datasets. Our code is available at https://github.com/HolmesShuan/Location-aware-Upsampling-for-Semantic-Segmentation

Qiang Chen, Anda Cheng, Xiangyu He et al.

Object location is fundamental to panoptic segmentation as it is related to all things and stuff in the image scene. Knowing the locations of objects in the image provides clues for segmenting and helps the network better understand the scene. How to integrate object location in both thing and stuff segmentation is a crucial problem. In this paper, we propose spatial information flows to achieve this objective. The flows can bridge all sub-tasks in panoptic segmentation by delivering the object's spatial context from the box regression task to others. More importantly, we design four parallel sub-networks to get a preferable adaptation of object spatial information in sub-tasks. Upon the sub-networks and the flows, we present a location-aware and unified framework for panoptic segmentation, denoted as SpatialFlow. We perform a detailed ablation study on each component and conduct extensive experiments to prove the effectiveness of SpatialFlow. Furthermore, we achieve state-of-the-art results, which are $47.9$ PQ and $62.5$ PQ respectively on MS-COCO and Cityscapes panoptic benchmarks. Code will be available at https://github.com/chensnathan/SpatialFlow.

Xiangyu He, Ke Cheng, Qiang Chen et al.

Long-range dependencies modeling, widely used in capturing spatiotemporal correlation, has shown to be effective in CNN dominated computer vision tasks. Yet neither stacks of convolutional operations to enlarge receptive fields nor recent nonlocal modules is computationally efficient. In this paper, we present a generic family of lightweight global descriptors for modeling the interactions between positions across different dimensions (e.g., channels, frames). This descriptor enables subsequent convolutions to access the informative global features with negligible computational complexity and parameters. Benchmark experiments show that the proposed method can complete state-of-the-art long-range mechanisms with a significant reduction in extra computing cost. Code available at https://github.com/HolmesShuan/Compact-Global-Descriptor.

Jiaqi Chen, Bang Zhang, Ruotian Ma et al.

Evaluating the step-by-step reliability of large language model (LLM) reasoning, such as Chain-of-Thought, remains challenging due to the difficulty and cost of obtaining high-quality step-level supervision. In this paper, we introduce Self-Play Critic (SPC), a novel approach where a critic model evolves its ability to assess reasoning steps through adversarial self-play games, eliminating the need for manual step-level annotation. SPC involves fine-tuning two copies of a base model to play two roles, namely a "sneaky generator" that deliberately produces erroneous steps designed to be difficult to detect, and a "critic" that analyzes the correctness of reasoning steps. These two models engage in an adversarial game in which the generator aims to fool the critic, while the critic model seeks to identify the generator's errors. Using reinforcement learning based on the game outcomes, the models iteratively improve; the winner of each confrontation receives a positive reward and the loser receives a negative reward, driving continuous self-evolution. Experiments on three reasoning process benchmarks (ProcessBench, PRM800K, DeltaBench) demonstrate that our SPC progressively enhances its error detection capabilities (e.g., accuracy increases from 70.8% to 77.7% on ProcessBench) and surpasses strong baselines, including distilled R1 model. Furthermore, SPC can guide the test-time search of diverse LLMs and significantly improve their mathematical reasoning performance on MATH500 and AIME2024, surpassing those guided by state-of-the-art process reward models.

Peisong Wang, Ruotian Ma, Bang Zhang et al.

Large language models (LLMs) excel at logical and algorithmic reasoning, yet their emotional intelligence (EQ) still lags far behind their cognitive prowess. While reinforcement learning from verifiable rewards (RLVR) has advanced in other domains, its application to dialogue-especially for emotional intelligence-remains underexplored. In this work, we introduce RLVER, the first end-to-end reinforcement learning framework that leverages verifiable emotion rewards from simulated users to cultivate higher-order empathetic abilities in LLMs. Within this framework, self-consistent affective simulated users engage in dialogue rollouts and produce deterministic emotion scores during conversations, serving as reward signals to guide the LLM's learning. Fine-tuning publicly available Qwen2.5-7B-Instruct model with PPO boosts its Sentient-Benchmark score from 13.3 to 79.2 while largely preserving mathematical and coding competence. Extensive experiments reveal that: (i) RLVER consistently improves multiple dialogue capabilities; (ii) Thinking and non-thinking models show distinct trends--thinking models excel in empathy and insight, while non-thinking models favor action; (iii) GRPO often yields stable gains, while PPO can push certain capabilities to a higher ceiling; (iv) More challenging environments are not always better-moderate ones can yield stronger outcomes. Our results show that RLVER is a practical route toward emotionally intelligent and broadly capable language agents.

Yuanteng Chen, Yuantian Shao, Peisong Wang et al.

Mixture-of-Experts (MoE) has demonstrated promising potential in scaling LLMs. However, it is hindered by two critical challenges: (1) substantial GPU memory consumption to load all experts; (2) low activated parameters cannot be equivalently translated into inference acceleration effects. In this work, we propose EAC-MoE, an Expert-Selection Aware Compressor for MoE-LLMs, which deeply aligns with the characteristics of MoE from the perspectives of quantization and pruning, and introduces two modules to address these two challenges respectively: (1) The expert selection bias caused by low-bit quantization is a major factor contributing to the performance degradation in MoE-LLMs. Based on this, we propose Quantization with Expert-Selection Calibration (QESC), which mitigates the expert selection bias by calibrating the routers within the MoE; (2) There are always certain experts that are not crucial for the corresponding tasks, yet causing inference latency. Therefore, we propose Pruning based on Expert-Selection Frequency (PESF), which significantly improves inference speed by pruning less frequently used experts for current task. Extensive experiments demonstrate that our approach significantly reduces memory usage and improves inference speed with minimal performance degradation.

Jiahe Qian, Yuhao Shen, Zhangtianyi Chen et al.

Medical visual question answering could support clinical decision making, yet current systems often fail under domain shift and produce answers that are weakly grounded in image evidence. This reliability gap arises when models attend to spurious regions and when retraining or additional labels are impractical at deployment time. We address this setting with CoTBox-TTT, an evidence-first test-time training approach that adapts a vision-language model at inference while keeping all backbones frozen. The method updates only a small set of continuous soft prompts. It identifies question-relevant regions through a visual chain-of-thought signal and encourages answer consistency across the original image and a localized crop. The procedure is label free, and plug and play with diverse backbones. Experiments on medical VQA show that the approach is practical for real deployments. For instance, adding CoTBox-TTT to LLaVA increases closed-ended accuracy by 12.3% on pathVQA.

Zhinan Xie, Peisong Wang, Shuang Qiu et al.

Speculative decoding has proven effective for accelerating inference in Large Language Models (LLMs), yet its extension to Vision-Language Models (VLMs) remains limited by the computational burden and semantic inconsistency introduced by visual tokens. Recent studies reveal that visual tokens in large VLMs are highly redundant, and most of them can be removed without compromising generation quality. Motivated by this observation, we propose HiViS (Hiding Visual Tokens from the Drafter for Speculative Decoding in Vision-Language Models), a framework that utilizes the target VLM as a semantic fusion model, allowing the drafter to obtain visual information without explicitly processing visual tokens, ensuring that the drafter's prefill sequence length matches that of the textual tokens. Furthermore, HiViS employs a time-step-aware aligned training scheme that allows the drafter to autonomously propagate and refine instructive visual-textual semantics during independent drafting, guided by step-dependent bias-correction residuals. Extensive experiments across representative VLMs and benchmarks demonstrate that HiViS achieves significant improvements in average acceptance length and speedup ratio.

Yuanteng Chen, Peisong Wang, Yuantian Shao et al.

Sparse Mixture-of-Experts (MoE) has become a key architecture for scaling large language models (LLMs) efficiently. Recent fine-grained MoE designs introduce hundreds of experts per layer, with multiple experts activated per token, enabling stronger specialization. However, during pre-training, routers are optimized mainly for stability and robustness: they converge prematurely and enforce balanced usage, limiting the full potential of model performance and efficiency at inference. In this work, we uncover two overlooked issues: (i) a few highly influential experts are underutilized due to premature and balanced routing decisions; and (ii) enforcing a fixed number of active experts per token introduces substantial redundancy. Instead of retraining models or redesigning MoE architectures, we introduce Ban&Pick, a post-training, plug-and-play strategy for smarter routing. Pick discovers and reinforces key experts-a small group with outsized impact on performance-leading to notable accuracy gains across domains. Ban further dynamically prunes redundant experts based on layer and token sensitivity, delivering faster inference with minimal accuracy loss. Experiments on fine-grained MoE-LLMs (DeepSeek, Qwen3) across math, code, and general reasoning benchmarks demonstrate that Ban\&Pick delivers free performance gains and inference acceleration without retraining or architectural changes. For instance, on Qwen3-30B-A3B, it improves accuracy from 80.67 to 84.66 on AIME2024 and from 65.66 to 68.18 on GPQA-Diamond, while accelerating inference by 1.25x under the vLLM.

Zhexin Li, Tong Yang, Peisong Wang et al.

In this paper, we propose a fully differentiable quantization method for vision transformer (ViT) named as Q-ViT, in which both of the quantization scales and bit-widths are learnable parameters. Specifically, based on our observation that heads in ViT display different quantization robustness, we leverage head-wise bit-width to squeeze the size of Q-ViT while preserving performance. In addition, we propose a novel technique named switchable scale to resolve the convergence problem in the joint training of quantization scales and bit-widths. In this way, Q-ViT pushes the limits of ViT quantization to 3-bit without heavy performance drop. Moreover, we analyze the quantization robustness of every architecture component of ViT and show that the Multi-head Self-Attention (MSA) and the Gaussian Error Linear Units (GELU) are the key aspects for ViT quantization. This study provides some insights for further research about ViT quantization. Extensive experiments on different ViT models, such as DeiT and Swin Transformer show the effectiveness of our quantization method. In particular, our method outperforms the state-of-the-art uniform quantization method by 1.5% on DeiT-Tiny.

Anda Cheng, Jiaxing Wang, Xi Sheryl Zhang et al.

Training deep neural networks (DNNs) for meaningful differential privacy (DP) guarantees severely degrades model utility. In this paper, we demonstrate that the architecture of DNNs has a significant impact on model utility in the context of private deep learning, whereas its effect is largely unexplored in previous studies. In light of this missing, we propose the very first framework that employs neural architecture search to automatic model design for private deep learning, dubbed as DPNAS. To integrate private learning with architecture search, we delicately design a novel search space and propose a DP-aware method for training candidate models. We empirically certify the effectiveness of the proposed framework. The searched model DPNASNet achieves state-of-the-art privacy/utility trade-offs, e.g., for the privacy budget of $(ε, δ)=(3, 1\times10^{-5})$, our model obtains test accuracy of $98.57\%$ on MNIST, $88.09\%$ on FashionMNIST, and $68.33\%$ on CIFAR-10. Furthermore, by studying the generated architectures, we provide several intriguing findings of designing private-learning-friendly DNNs, which can shed new light on model design for deep learning with differential privacy.

Weihan Chen, Peisong Wang, Jian Cheng

Quantization is a widely used technique to compress and accelerate deep neural networks. However, conventional quantization methods use the same bit-width for all (or most of) the layers, which often suffer significant accuracy degradation in the ultra-low precision regime and ignore the fact that emergent hardware accelerators begin to support mixed-precision computation. Consequently, we present a novel and principled framework to solve the mixed-precision quantization problem in this paper. Briefly speaking, we first formulate the mixed-precision quantization as a discrete constrained optimization problem. Then, to make the optimization tractable, we approximate the objective function with second-order Taylor expansion and propose an efficient approach to compute its Hessian matrix. Finally, based on the above simplification, we show that the original problem can be reformulated as a Multiple-Choice Knapsack Problem (MCKP) and propose a greedy search algorithm to solve it efficiently. Compared with existing mixed-precision quantization works, our method is derived in a principled way and much more computationally efficient. Moreover, extensive experiments conducted on the ImageNet dataset and various kinds of network architectures also demonstrate its superiority over existing uniform and mixed-precision quantization approaches.

Weixiang Xu, Qiang Chen, Xiangyu He et al.

Binary Neural Networks (BNNs) rely on a real-valued auxiliary variable W to help binary training. However, pioneering binary works only use W to accumulate gradient updates during backward propagation, which can not fully exploit its power and may hinder novel advances in BNNs. In this work, we explore the role of W in training besides acting as a latent variable. Notably, we propose to add W into the computation graph, making it perform as a real-valued feature extractor to aid the binary training. We make different attempts on how to utilize the real-valued weights and propose a specialized supervision. Visualization experiments qualitatively verify the effectiveness of our approach in making it easier to distinguish between different categories. Quantitative experiments show that our approach outperforms current state-of-the-arts, further closing the performance gap between floating-point networks and BNNs. Evaluation on ImageNet with ResNet-18 (Top-1 63.4%), ResNet-34 (Top-1 67.0%) achieves new state-of-the-art.

Xiangyu He, Qinghao Hu, Peisong Wang et al.

Convolutional neural networks are able to learn realistic image priors from numerous training samples in low-level image generation and restoration. We show that, for high-level image recognition tasks, we can further reconstruct "realistic" images of each category by leveraging intrinsic Batch Normalization (BN) statistics without any training data. Inspired by the popular VAE/GAN methods, we regard the zero-shot optimization process of synthetic images as generative modeling to match the distribution of BN statistics. The generated images serve as a calibration set for the following zero-shot network quantizations. Our method meets the needs for quantizing models based on sensitive information, \textit{e.g.,} due to privacy concerns, no data is available. Extensive experiments on benchmark datasets show that, with the help of generated data, our approach consistently outperforms existing data-free quantization methods.

Fanrong Li, Zitao Mo, Peisong Wang et al.

Object detection has made impressive progress in recent years with the help of deep learning. However, state-of-the-art algorithms are both computation and memory intensive. Though many lightweight networks are developed for a trade-off between accuracy and efficiency, it is still a challenge to make it practical on an embedded device. In this paper, we present a system-level solution for efficient object detection on a heterogeneous embedded device. The detection network is quantized to low bits and allows efficient implementation with shift operators. In order to make the most of the benefits of low-bit quantization, we design a dedicated accelerator with programmable logic. Inside the accelerator, a hybrid dataflow is exploited according to the heterogeneous property of different convolutional layers. We adopt a straightforward but resource-friendly column-prior tiling strategy to map the computation-intensive convolutional layers to the accelerator that can support arbitrary feature size. Other operations can be performed on the low-power CPU cores, and the entire system is executed in a pipelined manner. As a case study, we evaluate our object detection system on a real-world surveillance video with input size of 512x512, and it turns out that the system can achieve an inference speed of 18 fps at the cost of 6.9W (with display) with an mAP of 66.4 verified on the PASCAL VOC 2012 dataset.

Qinghao Hu, Peisong Wang, Jian Cheng

Deep convolutional neural networks (CNNs) have shown appealing performance on various computer vision tasks in recent years. This motivates people to deploy CNNs to realworld applications. However, most of state-of-art CNNs require large memory and computational resources, which hinders the deployment on mobile devices. Recent studies show that low-bit weight representation can reduce much storage and memory demand, and also can achieve efficient network inference. To achieve this goal, we propose a novel approach named BWNH to train Binary Weight Networks via Hashing. In this paper, we first reveal the strong connection between inner-product preserving hashing and binary weight networks, and show that training binary weight networks can be intrinsically regarded as a hashing problem. Based on this perspective, we propose an alternating optimization method to learn the hash codes instead of directly learning binary weights. Extensive experiments on CIFAR10, CIFAR100 and ImageNet demonstrate that our proposed BWNH outperforms current state-of-art by a large margin.

Jian Cheng, Peisong Wang, Gang Li et al.

Deep neural networks have evolved remarkably over the past few years and they are currently the fundamental tools of many intelligent systems. At the same time, the computational complexity and resource consumption of these networks also continue to increase. This will pose a significant challenge to the deployment of such networks, especially in real-time applications or on resource-limited devices. Thus, network acceleration has become a hot topic within the deep learning community. As for hardware implementation of deep neural networks, a batch of accelerators based on FPGA/ASIC have been proposed in recent years. In this paper, we provide a comprehensive survey of recent advances in network acceleration, compression and accelerator design from both algorithm and hardware points of view. Specifically, we provide a thorough analysis of each of the following topics: network pruning, low-rank approximation, network quantization, teacher-student networks, compact network design and hardware accelerators. Finally, we will introduce and discuss a few possible future directions.

Peisong Wang, Jian Cheng

In recent years, Deep Neural Networks (DNN) based methods have achieved remarkable performance in a wide range of tasks and have been among the most powerful and widely used techniques in computer vision. However, DNN-based methods are both computational-intensive and resource-consuming, which hinders the application of these methods on embedded systems like smart phones. To alleviate this problem, we introduce a novel Fixed-point Factorized Networks (FFN) for pretrained models to reduce the computational complexity as well as the storage requirement of networks. The resulting networks have only weights of -1, 0 and 1, which significantly eliminates the most resource-consuming multiply-accumulate operations (MACs). Extensive experiments on large-scale ImageNet classification task show the proposed FFN only requires one-thousandth of multiply operations with comparable accuracy.