Yiming Pan

GLM-5 Team, Aohan Zeng, Xin Lv et al. · tsinghua

We present GLM-5, a next-generation foundation model designed to transition the paradigm of vibe coding to agentic engineering. Building upon the agentic, reasoning, and coding (ARC) capabilities of its predecessor, GLM-5 adopts DSA to significantly reduce training and inference costs while maintaining long-context fidelity. To advance model alignment and autonomy, we implement a new asynchronous reinforcement learning infrastructure that drastically improves post-training efficiency by decoupling generation from training. Furthermore, we propose novel asynchronous agent RL algorithms that further improve RL quality, enabling the model to learn from complex, long-horizon interactions more effectively. Through these innovations, GLM-5 achieves state-of-the-art performance on major open benchmarks. Most critically, GLM-5 demonstrates unprecedented capability in real-world coding tasks, surpassing previous baselines in handling end-to-end software engineering challenges. Code, models, and more information are available at https://github.com/zai-org/GLM-5.

Zhiwei Fan, Yiming Pan, Daniel Coca

Coherent nonlinear wave dynamics are often strongly shaped by a compact set of physically meaningful descriptors of the initial state. Traditional neural operators typically treat the input-output mapping as a largely black-box high-dimensional regression problem, without explicitly exploiting this structured physical context. Common feature-integration strategies usually rely on direct concatenation or FiLM-style affine modulation in hidden latent spaces. Here we introduce a different paradigm, loosely inspired by the complementary roles of state evolution and physically meaningful observables in quantum mechanics: the wave field is learned through a standard DeepONet state pathway, while compact physical descriptors follow a parallel conditioning pathway and act as residual modulation factors on the state prediction. Based on this idea, we develop a Multi-Head Residual-Gated DeepONet (MH-RG), which combines a pre-branch residual modulator, a branch residual gate, and a trunk residual gate with a low-rank multi-head mechanism to capture multiple complementary conditioned response patterns without prohibitive parameter growth. We evaluate the framework on representative benchmarks including highly nonlinear conservative wave dynamics and dissipative trapped dynamics and further perform detailed mechanistic analyses of the learned multi-head gating behavior. Compared with feature-augmented baselines, MH-RG DeepONet achieves consistently lower error while better preserving phase coherence and the fidelity of physically relevant dynamical quantities.

GLM-4. 5 Team, Aohan Zeng, Xin Lv et al.

We present GLM-4.5, an open-source Mixture-of-Experts (MoE) large language model with 355B total parameters and 32B activated parameters, featuring a hybrid reasoning method that supports both thinking and direct response modes. Through multi-stage training on 23T tokens and comprehensive post-training with expert model iteration and reinforcement learning, GLM-4.5 achieves strong performance across agentic, reasoning, and coding (ARC) tasks, scoring 70.1% on TAU-Bench, 91.0% on AIME 24, and 64.2% on SWE-bench Verified. With much fewer parameters than several competitors, GLM-4.5 ranks 3rd overall among all evaluated models and 2nd on agentic benchmarks. We release both GLM-4.5 (355B parameters) and a compact version, GLM-4.5-Air (106B parameters), to advance research in reasoning and agentic AI systems. Code, models, and more information are available at https://github.com/zai-org/GLM-4.5.

Yiming Pan, Chengwei Hu, Xuancheng Huang et al.

Large language models (LLMs) have demonstrated strong potential in agentic tasks, particularly in slide generation. However, slide generation poses a fundamental challenge: the generation process is text-centric, whereas its quality is governed by visual aesthetics. This modality gap leads current models to frequently produce slides with aesthetically suboptimal layouts. Existing solutions typically rely either on heavy visual reflection, which incurs high inference cost yet yields limited gains; or on fine-tuning with large-scale datasets, which still provides weak and indirect aesthetic supervision. In contrast, the explicit use of aesthetic principles as supervision remains unexplored. In this work, we present AeSlides, a reinforcement learning framework with verifiable rewards for Aesthetic layout supervision in Slide generation. We introduce a suite of meticulously designed verifiable metrics to quantify slide layout quality, capturing key layout issues in an accurate, efficient, and low-cost manner. Leveraging these verifiable metrics, we develop a GRPO-based reinforcement learning method that directly optimizes slide generation models for aesthetically coherent layouts. With only 5K training prompts on GLM-4.7-Flash, AeSlides improves aspect ratio compliance from 36% to 85%, while reducing whitespace by 44%, element collisions by 43%, and visual imbalance by 28%. Human evaluation further shows a substantial improvement in overall quality, increasing scores from 3.31 to 3.56 (+7.6%), outperforming both model-based reward optimization and reflection-based agentic approaches, and even edging out Claude-Sonnet-4.5. These results demonstrate that such a verifiable aesthetic paradigm provides an efficient and scalable approach to aligning slide generation with human aesthetic preferences. Our repository is available at https://github.com/ympan0508/aeslides.

Yuan Zhu, Yanqiang Wang, Yadong An et al.

This paper focuses on a real-time vehicle detection and urban traffic behavior analysis system based on Unmanned Aerial Vehicle (UAV) traffic video. By using UAV to collect traffic data and combining the YOLOv8 model and SORT tracking algorithm, the object detection and tracking functions are implemented on the iOS mobile platform. For the problem of traffic data acquisition and analysis, the dynamic computing method is used to process the performance in real time and calculate the micro and macro traffic parameters of the vehicles, and real-time traffic behavior analysis is conducted and visualized. The experiment results reveals that the vehicle object detection can reach 98.27% precision rate and 87.93% recall rate, and the real-time processing capacity is stable at 30 frames per seconds. This work integrates drone technology, iOS development, and deep learning techniques to integrate traffic video acquisition, object detection, object tracking, and traffic behavior analysis functions on mobile devices. It provides new possibilities for lightweight traffic information collection and data analysis, and offers innovative solutions to improve the efficiency of analyzing road traffic conditions and addressing transportation issues for transportation authorities.

Fuheng Zhao, Jiayue Chen, Yiming Pan et al.

We present the LLM ORDER BY operator as a logical abstraction and study its physical implementations within a unified evaluation framework. Our experiments show that no single approach is universally optimal, with effectiveness depending on query characteristics and data. We introduce three new designs: an agreement-based batch-size policy, a majority voting mechanism for pairwise sorting, and a two-way external merge sort adapted for LLMs. With extensive experiments, our agreement-based procedure is effective at determining batch size for value-based methods, the majority-voting mechanism consistently strengthens pairwise comparisons on GPT-4o, and external merge sort achieves high accuracy-efficiency trade-offs across datasets and models. We further observe a log-linear scaling between compute cost and ordering quality, offering the first step toward principled cost models for LLM powered data systems.

Weiyang Geng, Yiming Pan, Zhecong Xing et al.

This study proposes a hybrid model based on Transformers, named MSCMHMST, aimed at addressing key challenges in traffic flow prediction. Traditional single-method approaches show limitations in traffic prediction tasks, whereas hybrid methods, by integrating the strengths of different models, can provide more accurate and robust predictions. The MSCMHMST model introduces a multi-head, multi-scale attention mechanism, allowing the model to parallel process different parts of the data and learn its intrinsic representations from multiple perspectives, thereby enhancing the model's ability to handle complex situations. This mechanism enables the model to capture features at various scales effectively, understanding both short-term changes and long-term trends. Verified through experiments on the PeMS04/08 dataset with specific experimental settings, the MSCMHMST model demonstrated excellent robustness and accuracy in long, medium, and short-term traffic flow predictions. The results indicate that this model has significant potential, offering a new and effective solution for the field of traffic flow prediction.