Wei Qin

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.

Yunshi Lan, Xiang Li, Xin Liu et al.

Zero-shot Visual Question Answering (VQA) is a prominent vision-language task that examines both the visual and textual understanding capability of systems in the absence of training data. Recently, by converting the images into captions, information across multi-modalities is bridged and Large Language Models (LLMs) can apply their strong zero-shot generalization capability to unseen questions. To design ideal prompts for solving VQA via LLMs, several studies have explored different strategies to select or generate question-answer pairs as the exemplar prompts, which guide LLMs to answer the current questions effectively. However, they totally ignore the role of question prompts. The original questions in VQA tasks usually encounter ellipses and ambiguity which require intermediate reasoning. To this end, we present Reasoning Question Prompts for VQA tasks, which can further activate the potential of LLMs in zero-shot scenarios. Specifically, for each question, we first generate self-contained questions as reasoning question prompts via an unsupervised question edition module considering sentence fluency, semantic integrity and syntactic invariance. Each reasoning question prompt clearly indicates the intent of the original question. This results in a set of candidate answers. Then, the candidate answers associated with their confidence scores acting as answer heuristics are fed into LLMs and produce the final answer. We evaluate reasoning question prompts on three VQA challenges, experimental results demonstrate that they can significantly improve the results of LLMs on zero-shot setting and outperform existing state-of-the-art zero-shot methods on three out of four data sets. Our source code is publicly released at \url{https://github.com/ECNU-DASE-NLP/RQP}.

Weijieying Ren, Xinlong Li, Lei Wang et al.

Existing research has shown that large language models (LLMs) exhibit remarkable performance in language understanding and generation. However, when LLMs are continuously fine-tuned on complex and diverse domain-specific downstream tasks, the inference performance on historical tasks decreases dramatically, which is known as a catastrophic forgetting problem. A trade-off needs to be kept between learning plasticity and memory stability. Plenty of existing works have explored strategies like memory replay, regularization and parameter isolation, but little is known about the geometric connection of various adjacent minima in the continual LLMs fine-tuning scenarios. In this work, we investigate the geometric connections of different minima through the lens of mode connectivity, which means different minima can be connected by a low-loss valley. Through extensive experiments, we uncover the mode connectivity phenomenon in the LLMs continual learning scenario and find that it can strike a balance between plasticity and stability. Building upon these findings, we propose a simple yet effective method called Interpolation-based LoRA (I-LoRA), which constructs a dual-memory experience replay framework based on LoRA parameter interpolations. Extensive experiments and analysis on eight domain-specific CL benchmarks demonstrate that I-LoRA consistently show significant improvement over the previous state-of-the-art approaches with up to $11\%$ performance gains, providing a strong baseline and insights for future research on the large language model continual learning problem. Our code is available at \url{https://github.com/which47/LLMCL}.

Junhao Tan, Songwen Pei, Wei Qin et al.

Frequency information (e.g., Discrete Wavelet Transform and Fast Fourier Transform) has been widely applied to solve the issue of Low-Light Image Enhancement (LLIE). However, existing frequency-based models primarily operate in the simple wavelet or Fourier space of images, which lacks utilization of valid global and local information in each space. We found that wavelet frequency information is more sensitive to global brightness due to its low-frequency component while Fourier frequency information is more sensitive to local details due to its phase component. In order to achieve superior preliminary brightness enhancement by optimally integrating spatial channel information with low-frequency components in the wavelet transform, we introduce channel-wise Mamba, which compensates for the long-range dependencies of CNNs and has lower complexity compared to Diffusion and Transformer models. So in this work, we propose a novel Wavelet-based Mamba with Fourier Adjustment model called WalMaFa, consisting of a Wavelet-based Mamba Block (WMB) and a Fast Fourier Adjustment Block (FFAB). We employ an Encoder-Latent-Decoder structure to accomplish the end-to-end transformation. Specifically, WMB is adopted in the Encoder and Decoder to enhance global brightness while FFAB is adopted in the Latent to fine-tune local texture details and alleviate ambiguity. Extensive experiments demonstrate that our proposed WalMaFa achieves state-of-the-art performance with fewer computational resources and faster speed. Code is now available at: https://github.com/mcpaulgeorge/WalMaFa.

Qinghua Shen, Weijieying Ren, Wei Qin

Continual graph learning (CGL) studies the problem of learning from an infinite stream of graph data, consolidating historical knowledge, and generalizing it to the future task. At once, only current graph data are available. Although some recent attempts have been made to handle this task, we still face two potential challenges: 1) most of existing works only manipulate on the intermediate graph embedding and ignore intrinsic properties of graphs. It is non-trivial to differentiate the transferred information across graphs. 2) recent attempts take a parameter-sharing policy to transfer knowledge across time steps or progressively expand new architecture given shifted graph distribution. Learning a single model could loss discriminative information for each graph task while the model expansion scheme suffers from high model complexity. In this paper, we point out that latent relations behind graph edges can be attributed as an invariant factor for the evolving graphs and the statistical information of latent relations evolves. Motivated by this, we design a relation-aware adaptive model, dubbed as RAM-CG, that consists of a relation-discovery modular to explore latent relations behind edges and a task-awareness masking classifier to accounts for the shifted. Extensive experiments show that RAM-CG provides significant 2.2%, 6.9% and 6.6% accuracy improvements over the state-of-the-art results on CitationNet, OGBN-arxiv and TWITCH dataset, respective.

Weijieying Ren, Tianxiang Zhao, Wei Qin et al.

In many real-world scenarios, distribution shifts exist in the streaming data across time steps. Many complex sequential data can be effectively divided into distinct regimes that exhibit persistent dynamics. Discovering the shifted behaviors and the evolving patterns underlying the streaming data are important to understand the dynamic system. Existing methods typically train one robust model to work for the evolving data of distinct distributions or sequentially adapt the model utilizing explicitly given regime boundaries. However, there are two challenges: (1) shifts in data streams could happen drastically and abruptly without precursors. Boundaries of distribution shifts are usually unavailable, and (2) training a shared model for all domains could fail to capture varying patterns. This paper aims to solve the problem of sequential data modeling in the presence of sudden distribution shifts that occur without any precursors. Specifically, we design a Bayesian framework, dubbed as T-SaS, with a discrete distribution-modeling variable to capture abrupt shifts of data. Then, we design a model that enable adaptation with dynamic network selection conditioned on that discrete variable. The proposed method learns specific model parameters for each distribution by learning which neurons should be activated in the full network. A dynamic masking strategy is adopted here to support inter-distribution transfer through the overlapping of a set of sparse networks. Extensive experiments show that our proposed method is superior in both accurately detecting shift boundaries to get segments of varying distributions and effectively adapting to downstream forecast or classification tasks.

Jilong Liu, Pengyang Shao, Wei Qin et al.

Knowledge Graph Question Answering (KGQA) aims to improve factual accuracy by leveraging structured knowledge. However, real-world Knowledge Graphs (KGs) are often incomplete, leading to the problem of Incomplete KGQA (IKGQA). A common solution is to incorporate external data to fill knowledge gaps, but existing methods lack the capacity to adaptively and contextually fuse multiple sources, failing to fully exploit their complementary strengths. To this end, we propose Debate over Mixed-knowledge (DoM), a novel framework that enables dynamic integration of structured and unstructured knowledge for IKGQA. Built upon the Multi-Agent Debate paradigm, DoM assigns specialized agents to perform inference over knowledge graphs and external texts separately, and coordinates their outputs through iterative interaction. It decomposes the input question into sub-questions, retrieves evidence via dual agents (KG and Retrieval-Augmented Generation, RAG), and employs a judge agent to evaluate and aggregate intermediate answers. This collaboration exploits knowledge complementarity and enhances robustness to KG incompleteness. In addition, existing IKGQA datasets simulate incompleteness by randomly removing triples, failing to capture the irregular and unpredictable nature of real-world knowledge incompleteness. To address this, we introduce a new dataset, Incomplete Knowledge Graph WebQuestions, constructed by leveraging real-world knowledge updates. These updates reflect knowledge beyond the static scope of KGs, yielding a more realistic and challenging benchmark. Through extensive experiments, we show that DoM consistently outperforms state-of-the-art baselines.

Lubian Bai, Xiuyuan Zhang, Siqi Zhang et al.

Integrating ground-level geospatial data with rich geographic context, like OpenStreetMap (OSM), into remote sensing (RS) foundation models (FMs) is essential for advancing geospatial intelligence and supporting a broad spectrum of tasks. However, modality gap between RS and OSM data, including differences in data structure, content, and spatial granularity, makes effective synergy highly challenging, and most existing RS FMs focus on imagery alone. To this end, this study presents GeoLink, a multimodal framework that leverages OSM data to enhance RS FM during both the pretraining and downstream task stages. Specifically, GeoLink enhances RS self-supervised pretraining using multi-granularity learning signals derived from OSM data, guided by cross-modal spatial correlations for information interaction and collaboration. It also introduces image mask-reconstruction to enable sparse input for efficient pretraining. For downstream tasks, GeoLink generates both unimodal and multimodal fine-grained encodings to support a wide range of applications, from common RS interpretation tasks like land cover classification to more comprehensive geographic tasks like urban function zone mapping. Extensive experiments show that incorporating OSM data during pretraining enhances the performance of the RS image encoder, while fusing RS and OSM data in downstream tasks improves the FM's adaptability to complex geographic scenarios. These results underscore the potential of multimodal synergy in advancing high-level geospatial artificial intelligence. Moreover, we find that spatial correlation plays a crucial role in enabling effective multimodal geospatial data integration. Code, checkpoints, and using examples are released at https://github.com/bailubin/GeoLink_NeurIPS2025

Fan Zhang, Wei Qin, Weijieying Ren et al.

In the real-world setting, data often follows a long-tailed distribution, where head classes contain significantly more training samples than tail classes. Consequently, models trained on such data tend to be biased toward head classes. The medium of this bias is imbalanced gradients, which include not only the ratio of scale between positive and negative gradients but also imbalanced gradients from different negative classes. Therefore, we propose the Gradient-Aware Logit Adjustment (GALA) loss, which adjusts the logits based on accumulated gradients to balance the optimization process. Additionally, We find that most of the solutions to long-tailed problems are still biased towards head classes in the end, and we propose a simple and post hoc prediction re-balancing strategy to further mitigate the basis toward head class. Extensive experiments are conducted on multiple popular long-tailed recognition benchmark datasets to evaluate the effectiveness of these two designs. Our approach achieves top-1 accuracy of 48.5\%, 41.4\%, and 73.3\% on CIFAR100-LT, Places-LT, and iNaturalist, outperforming the state-of-the-art method GCL by a significant margin of 3.62\%, 0.76\% and 1.2\%, respectively. Code is available at https://github.com/lt-project-repository/lt-project.

Zhenxing Cheng, Peng Wang, Yue Liu et al.

Power capacitor device is a widely used reactive power compensation equipment in power transmission and distribution system which can easily have internal fault and therefore affects the safe operation of the power system. An intelligent non-destructive testing (I-NDT) method based on ICT is proposed to test the quality of power capacitors automatically in this study. The internal structure of power capacitors would be scanned by the ICT device and then defects could be recognized by the SSD algorithm. Moreover, the data data augmentation algorithm is used to extend the image set to improve the stability and accuracy of the trained SSD model.

Jilong Liu, Yonghui Yang, Pengyang Shao et al.

Direct Preference Optimization (DPO) aligns large language models by optimizing pairwise preferences and has shown remarkable effectiveness as a simple and scalable alternative to RLHF. However, in practice, preference data are often noisy. Existing robust variants of DPO mainly rely on uniform objective modifications or global reweighting. While partially effective, these methods treat noisy samples as a homogeneous source of uncertainty and fail to distinguish between different noise types, leading to sub-optimal alignment robustness. In this work, we show that robust preference alignment benefits from addressing different noise types with targeted interventions rather than uniform regularization. We propose winsorized Direct Preference Optimization~(wDPO), a robust LLM alignment approach with hierarchical winsorization. Specifically, wDPO adopts a reward-free hierarchical intervention strategy that leverages only signals already available during DPO training. It first uses the implicit margin from DPO log-ratio to identify heterogeneous noise patterns without relying on external reward models. For hard noise, wDPO performs a data-level intervention by sparsely correcting strongly inconsistent preference pairs. For ambiguous comparisons, it applies a gradient-level intervention through soft winsorization, capping extreme losses in the high-loss tail to prevent weakly informative samples from dominating gradient updates. Extensive experiments on PKU-SafeRLHF and multiple external safety benchmarks demonstrate that wDPO consistently improves preference alignment quality and robustness over vanilla DPO and strong DPO-family baselines, with particularly pronounced gains under controlled label-flip noise.

Xiabao Wu, Yongchao Liu, Wei Qin et al.

Graph neural networks (GNNs) have delivered remarkable results in various fields. However, the rapid increase in the scale of graph data has introduced significant performance bottlenecks for GNN inference. Both computational complexity and memory usage have risen dramatically, with memory becoming a critical limitation. Although graph sampling-based subgraph learning methods can help mitigate computational and memory demands, they come with drawbacks such as information loss and high redundant computation among subgraphs. This paper introduces an innovative processing paradgim for distributed graph learning that abstracts GNNs with a new set of programming interfaces and leverages Just-In-Time (JIT) compilation technology to its full potential. This paradigm enables GNNs to highly exploit the computational resources of distributed clusters by eliminating the drawbacks of subgraph learning methods, leading to a more efficient inference process. Our experimental results demonstrate that on industry-scale graphs of up to \textbf{500 million nodes and 22.4 billion edges}, our method can produce a performance boost of up to \textbf{27.4 times}.

Wei Qin, Zetong Chen, Lei Wang et al.

This paper proposes a new depression detection system based on LLMs that is both interpretable and interactive. It not only provides a diagnosis, but also diagnostic evidence and personalized recommendations based on natural language dialogue with the user. We address challenges such as the processing of large amounts of text and integrate professional diagnostic criteria. Our system outperforms traditional methods across various settings and is demonstrated through case studies.

Zhenwen Liang, Jipeng Zhang, Lei Wang et al.

Math word problem (MWP) solving faces a dilemma in number representation learning. In order to avoid the number representation issue and reduce the search space of feasible solutions, existing works striving for MWP solving usually replace real numbers with symbolic placeholders to focus on logic reasoning. However, different from common symbolic reasoning tasks like program synthesis and knowledge graph reasoning, MWP solving has extra requirements in numerical reasoning. In other words, instead of the number value itself, it is the reusable numerical property that matters more in numerical reasoning. Therefore, we argue that injecting numerical properties into symbolic placeholders with contextualized representation learning schema can provide a way out of the dilemma in the number representation issue here. In this work, we introduce this idea to the popular pre-training language model (PLM) techniques and build MWP-BERT, an effective contextual number representation PLM. We demonstrate the effectiveness of our MWP-BERT on MWP solving and several MWP-specific understanding tasks on both English and Chinese benchmarks.