Zhiwei Liang

Minwei Kong, Chonghe Jiang, Ao Qu et al.

Large language models (LLMs) are increasingly used for optimization modeling and solver-code generation, yet practical operations research and optimization problems often require a harder capability: designing scalable algorithms that exploit problem structure and outperform direct formulation-and-solve baselines. Existing benchmarks are limited to small or simplified examples far below real-world scale and complexity. We introduce FrontierOR, among the first benchmarks to systematically evaluate LLM-based efficient algorithm design for realistic large-scale optimization problems. FrontierOR includes 180 tasks derived from methodologically diverse papers published in top-tier operations research venues, each with standardized instances and a hidden, expert-verified evaluation suite. We evaluate seven LLMs spanning frontier, cost-effective, and open-source models both in one-shot and test-time evolution settings. The results reveal that frontier models still struggle to move from executable formulations to efficient optimization algorithms: the strongest one-shot model outperforms Gurobi in only 31% of cases in both solution quality and computational efficiency, and even strong coding agents with test-time evolution achieve only 50% on selected hard tasks. FrontierOR establishes a practical evaluation platform for LLM-based optimization algorithm design, which enables future LLMs and agents to be systematically tested on whether they can move beyond correct formulation toward a feasible, high-quality, and efficient algorithm. Our FrontierOR Benchmark is available at https://anonymous.4open.science/r/efficient-opt-bench-F03D.

Qingxiao Sun, Tongxuan Liu, Shen Zhang et al.

Recommendation system delivers substantial economic benefits by providing personalized predictions. Generative recommendation (GR) integrates LLMs to enhance the understanding of long user-item sequences. Despite employing attention-based architectures, GR's workload differs markedly from that of LLM serving. GR typically processes long prompt while producing short, fixed-length outputs, yet the computational cost of each decode phase is especially high due to the large beam width. In addition, since the beam search involves a vast item space, the sorting overhead becomes particularly time-consuming. We propose xGR, a GR-oriented serving system that meets strict low-latency requirements under highconcurrency scenarios. First, xGR unifies the processing of prefill and decode phases through staged computation and separated KV cache. Second, xGR enables early sorting termination and mask-based item filtering with data structure reuse. Third, xGR reconstructs the overall pipeline to exploit multilevel overlap and multi-stream parallelism. Our experiments with real-world recommendation service datasets demonstrate that xGR achieves at least 3.49x throughput compared to the state-of-the-art baseline under strict latency constraints.

Tongxuan Liu, Tao Peng, Peijun Yang et al.

We introduce xLLM, an intelligent and efficient Large Language Model (LLM) inference framework designed for high-performance, large-scale enterprise-grade serving, with deep optimizations for diverse AI accelerators. To address these challenges, xLLM builds a novel decoupled service-engine architecture. At the service layer, xLLM-Service features an intelligent scheduling module that efficiently processes multimodal requests and co-locates online and offline tasks through unified elastic scheduling to maximize cluster utilization. This module also relies on a workload-adaptive dynamic Prefill-Decode (PD) disaggregation policy and a novel Encode-Prefill-Decode (EPD) disaggregation policy designed for multimodal inputs. Furthermore, it incorporates a distributed architecture to provide global KV Cache management and robust fault-tolerant capabilities for high availability. At the engine layer, xLLM-Engine co-optimizes system and algorithm designs to fully saturate computing resources. This is achieved through comprehensive multi-layer execution pipeline optimizations, an adaptive graph mode and an xTensor memory management. xLLM-Engine also further integrates algorithmic enhancements such as optimized speculative decoding and dynamic EPLB, collectively serving to substantially boost throughput and inference efficiency. Extensive evaluations demonstrate that xLLM delivers significantly superior performance and resource efficiency. Under identical TPOT constraints, xLLM achieves throughput up to 1.7x that of MindIE and 2.2x that of vLLM-Ascend with Qwen-series models, while maintaining an average throughput of 1.7x that of MindIE with Deepseek-series models. xLLM framework is publicly available at https://github.com/jd-opensource/xllm and https://github.com/jd-opensource/xllm-service.

Jiawei Shi, Songhao Zhu, Zhiwei Liang

Facial Expression Recognition (FER) is a classification task that points to face variants. Hence, there are certain affinity features between facial expressions, receiving little attention in the FER literature. Convolution padding, despite helping capture the edge information, causes erosion of the feature map simultaneously. After multi-layer filling convolution, the output feature map named albino feature definitely weakens the representation of the expression. To tackle these challenges, we propose a novel architecture named Amending Representation Module (ARM). ARM is a substitute for the pooling layer. Theoretically, it can be embedded in the back end of any network to deal with the Padding Erosion. ARM efficiently enhances facial expression representation from two different directions: 1) reducing the weight of eroded features to offset the side effect of padding, and 2) decomposing facial features to simplify representation learning. Experiments on public benchmarks prove that our ARM boosts the performance of FER remarkably. The validation accuracies are respectively 90.42% on RAF-DB, 65.2% on Affect-Net, and 58.71% on SFEW, exceeding current state-of-the-art methods. Our implementation and trained models are available at https://github.com/JiaweiShiCV/Amend-Representation-Module.

James Wen, Sahil Nalawade, Zhiwei Liang et al. · deepmind, harvard

Background: Generative artificial intelligence (AI) deployment in academic medical settings raises copyright compliance concerns. Dana-Farber Cancer Institute implemented GPT4DFCI, an internal generative AI tool utilizing OpenAI models, that is approved for enterprise use in research and operations. Given (1) the exceptionally broad adoption of the tool in our organization, (2) our research mission, and (3) the shared responsibility model required to benefit from Customer Copyright Commitment in Azure OpenAI Service products, we deemed rigorous copyright compliance testing necessary. Case Description: We conducted a structured red teaming exercise in Nov. 2024, with 42 participants from academic, industry, and government institutions. Four teams attempted to extract copyrighted content from GPT4DFCI across four domains: literary works, news articles, scientific publications, and access-restricted clinical notes. Teams successfully extracted verbatim book dedications and near-exact passages through various strategies. News article extraction failed despite jailbreak attempts. Scientific article reproduction yielded only high-level summaries. Clinical note testing revealed appropriate privacy safeguards. Discussion: The successful extraction of literary content indicates potential copyrighted material presence in training data, necessitating inference-time filtering. Differential success rates across content types suggest varying protective mechanisms. The event led to implementation of a copyright-specific meta-prompt in GPT4DFCI; this mitigation has been in production since Jan. 2025. Conclusion: Systematic red teaming revealed specific vulnerabilities in generative AI copyright compliance, leading to concrete mitigation strategies. Academic medical institutions deploying generative AI should implement continuous testing protocols to ensure legal and ethical compliance.

Yunbin Zhao, Songhao Zhu, Dongsheng Wang et al.

Occluded person re-identification is one of the challenging areas of computer vision, which faces problems such as inefficient feature representation and low recognition accuracy. Convolutional neural network pays more attention to the extraction of local features, therefore it is difficult to extract features of occluded pedestrians and the effect is not so satisfied. Recently, vision transformer is introduced into the field of re-identification and achieves the most advanced results by constructing the relationship of global features between patch sequences. However, the performance of vision transformer in extracting local features is inferior to that of convolutional neural network. Therefore, we design a partial feature transformer-based person re-identification framework named PFT. The proposed PFT utilizes three modules to enhance the efficiency of vision transformer. (1) Patch full dimension enhancement module. We design a learnable tensor with the same size as patch sequences, which is full-dimensional and deeply embedded in patch sequences to enrich the diversity of training samples. (2) Fusion and reconstruction module. We extract the less important part of obtained patch sequences, and fuse them with original patch sequence to reconstruct the original patch sequences. (3) Spatial Slicing Module. We slice and group patch sequences from spatial direction, which can effectively improve the short-range correlation of patch sequences. Experimental results over occluded and holistic re-identification datasets demonstrate that the proposed PFT network achieves superior performance consistently and outperforms the state-of-the-art methods.

Youwei Song, Jiahai Wang, Zhiwei Liang et al.

Aspect based sentiment analysis aims to identify the sentimental tendency towards a given aspect in text. Fine-tuning of pretrained BERT performs excellent on this task and achieves state-of-the-art performances. Existing BERT-based works only utilize the last output layer of BERT and ignore the semantic knowledge in the intermediate layers. This paper explores the potential of utilizing BERT intermediate layers to enhance the performance of fine-tuning of BERT. To the best of our knowledge, no existing work has been done on this research. To show the generality, we also apply this approach to a natural language inference task. Experimental results demonstrate the effectiveness and generality of the proposed approach.

Zhiyue Liu, Jiahai Wang, Zhiwei Liang

Generating multiple categories of texts is a challenging task and draws more and more attention. Since generative adversarial nets (GANs) have shown competitive results on general text generation, they are extended for category text generation in some previous works. However, the complicated model structures and learning strategies limit their performance and exacerbate the training instability. This paper proposes a category-aware GAN (CatGAN) which consists of an efficient category-aware model for category text generation and a hierarchical evolutionary learning algorithm for training our model. The category-aware model directly measures the gap between real samples and generated samples on each category, then reducing this gap will guide the model to generate high-quality category samples. The Gumbel-Softmax relaxation further frees our model from complicated learning strategies for updating CatGAN on discrete data. Moreover, only focusing on the sample quality normally leads the mode collapse problem, thus a hierarchical evolutionary learning algorithm is introduced to stabilize the training procedure and obtain the trade-off between quality and diversity while training CatGAN. Experimental results demonstrate that CatGAN outperforms most of the existing state-of-the-art methods.