Zongwei Li

Tencent Hunyuan Team, Ao Liu, Botong Zhou et al. · tencent-ai

As Large Language Models (LLMs) rapidly advance, we introduce Hunyuan-TurboS, a novel large hybrid Transformer-Mamba Mixture of Experts (MoE) model. It synergistically combines Mamba's long-sequence processing efficiency with Transformer's superior contextual understanding. Hunyuan-TurboS features an adaptive long-short chain-of-thought (CoT) mechanism, dynamically switching between rapid responses for simple queries and deep "thinking" modes for complex problems, optimizing computational resources. Architecturally, this 56B activated (560B total) parameter model employs 128 layers (Mamba2, Attention, FFN) with an innovative AMF/MF block pattern. Faster Mamba2 ensures linear complexity, Grouped-Query Attention minimizes KV cache, and FFNs use an MoE structure. Pre-trained on 16T high-quality tokens, it supports a 256K context length and is the first industry-deployed large-scale Mamba model. Our comprehensive post-training strategy enhances capabilities via Supervised Fine-Tuning (3M instructions), a novel Adaptive Long-short CoT Fusion method, Multi-round Deliberation Learning for iterative improvement, and a two-stage Large-scale Reinforcement Learning process targeting STEM and general instruction-following. Evaluations show strong performance: overall top 7 rank on LMSYS Chatbot Arena with a score of 1356, outperforming leading models like Gemini-2.0-Flash-001 (1352) and o4-mini-2025-04-16 (1345). TurboS also achieves an average of 77.9% across 23 automated benchmarks. Hunyuan-TurboS balances high performance and efficiency, offering substantial capabilities at lower inference costs than many reasoning models, establishing a new paradigm for efficient large-scale pre-trained models.

Zongwei Li, Lianghao Xia, Hua Hua et al.

Recent advances in Graph Neural Networks (GNNs) have revolutionized graph-structured data modeling, yet traditional GNNs struggle with complex heterogeneous structures prevalent in real-world scenarios. Despite progress in handling heterogeneous interactions, two fundamental challenges persist: noisy data significantly compromising embedding quality and learning performance, and existing methods' inability to capture intricate semantic transitions among heterogeneous relations, which impacts downstream predictions. To address these fundamental issues, we present the Heterogeneous Graph Diffusion Model (DiffGraph), a pioneering framework that introduces an innovative cross-view denoising strategy. This advanced approach transforms auxiliary heterogeneous data into target semantic spaces, enabling precise distillation of task-relevant information. At its core, DiffGraph features a sophisticated latent heterogeneous graph diffusion mechanism, implementing a novel forward and backward diffusion process for superior noise management. This methodology achieves simultaneous heterogeneous graph denoising and cross-type transition, while significantly simplifying graph generation through its latent-space diffusion capabilities. Through rigorous experimental validation on both public and industrial datasets, we demonstrate that DiffGraph consistently surpasses existing methods in link prediction and node classification tasks, establishing new benchmarks for robustness and efficiency in heterogeneous graph processing. The model implementation is publicly available at: https://github.com/HKUDS/DiffGraph.

Zongwei Li, Lianghao Xia, Chao Huang

Social recommendation has emerged as a powerful approach to enhance personalized recommendations by leveraging the social connections among users, such as following and friend relations observed in online social platforms. The fundamental assumption of social recommendation is that socially-connected users exhibit homophily in their preference patterns. This means that users connected by social ties tend to have similar tastes in user-item activities, such as rating and purchasing. However, this assumption is not always valid due to the presence of irrelevant and false social ties, which can contaminate user embeddings and adversely affect recommendation accuracy. To address this challenge, we propose a novel diffusion-based social denoising framework for recommendation (RecDiff). Our approach utilizes a simple yet effective hidden-space diffusion paradigm to alleivate the noisy effect in the compressed and dense representation space. By performing multi-step noise diffusion and removal, RecDiff possesses a robust ability to identify and eliminate noise from the encoded user representations, even when the noise levels vary. The diffusion module is optimized in a downstream task-aware manner, thereby maximizing its ability to enhance the recommendation process. We conducted extensive experiments to evaluate the efficacy of our framework, and the results demonstrate its superiority in terms of recommendation accuracy, training efficiency, and denoising effectiveness. The source code for the model implementation is publicly available at: https://github.com/HKUDS/RecDiff.