Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning

• URL: http://arxiv.org/abs/2510.19338v2
• Date: Thu, 23 Oct 2025 06:33:17 GMT
• Title: Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning
• Authors: Ling Team, Bin Han, Caizhi Tang, Chen Liang, Donghao Zhang, Fan Yuan, Feng Zhu, Jie Gao, Jingyu Hu, Longfei Li, Meng Li, Mingyang Zhang, Peijie Jiang, Peng Jiao, Qian Zhao, Qingyuan Yang, Wenbo Shen, Xinxing Yang, Yalin Zhang, Yankun Ren, Yao Zhao, Yibo Cao, Yixuan Sun, Yue Zhang, Yuchen Fang, Zibin Lin, Zixuan Cheng, Jun Zhou,
• Abstract summary: We present the Ring-linear model series, specifically including Ring-mini-linear-2.0 and Ring-flash-linear-2.0.<n>Both models adopt a hybrid architecture that effectively integrates linear attention and softmax attention.<n>Compared to a 32 billion parameter dense model, this series reduces inference cost to 1/10, and compared to the original Ring series, the cost is also reduced by over 50%.
• Score: 73.10669391954801
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: In this technical report, we present the Ring-linear model series, specifically including Ring-mini-linear-2.0 and Ring-flash-linear-2.0. Ring-mini-linear-2.0 comprises 16B parameters and 957M activations, while Ring-flash-linear-2.0 contains 104B parameters and 6.1B activations. Both models adopt a hybrid architecture that effectively integrates linear attention and softmax attention, significantly reducing I/O and computational overhead in long-context inference scenarios. Compared to a 32 billion parameter dense model, this series reduces inference cost to 1/10, and compared to the original Ring series, the cost is also reduced by over 50%. Furthermore, through systematic exploration of the ratio between different attention mechanisms in the hybrid architecture, we have identified the currently optimal model structure. Additionally, by leveraging our self-developed high-performance FP8 operator library-linghe, overall training efficiency has been improved by 50%. Benefiting from the high alignment between the training and inference engine operators, the models can undergo long-term, stable, and highly efficient optimization during the reinforcement learning phase, consistently maintaining SOTA performance across multiple challenging complex reasoning benchmarks.

Related papers

• MiniCPM-SALA: Hybridizing Sparse and Linear Attention for Efficient Long-Context Modeling [80.48332380100915]
  MiniCPM-SALA is a hybrid model that integrates the high-fidelity long-context modeling of sparse attention with the global efficiency of linear attention.<n>On a single NVIDIA A6000D GPU, the model achieves up to 3.5x the inference speed of the full-attention model at the sequence length of 256K tokens.
  arXiv  Detail & Related papers  (2026-02-12T09:37:05Z)
• Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model [100.86587937568832]
  Ring-1T is the first open-source, state-of-the-art thinking model with a trillion-scale parameter.<n>It features 1 trillion total parameters and activates approximately 50 billion per token.
  arXiv  Detail & Related papers  (2025-10-21T17:46:14Z)
• Causal-Guided Dimension Reduction for Efficient Pareto Optimization [2.9013001432962255]
  CaDRO builds a causal map through a hybrid observational-interventional process, ranking parameters by their causal effect on the objectives.<n>Low-impact parameters are fixed to values from high-quality solutions, while critical drivers remain active in the search.<n>Across amplifiers, regulators, and RF circuits, CaDRO converges up to 10$times$ faster than NSGA-II.
  arXiv  Detail & Related papers  (2025-10-11T00:41:04Z)
• Towards Greater Leverage: Scaling Laws for Efficient Mixture-of-Experts Language Models [25.608085561102566]
  We introduce Leverage Efficiency (EL), a metric quantifying the computational advantage of an MoE model over a dense equivalent.<n>EL is driven by the expert activation ratio and the total compute budget, both following predictable power laws.<n>We integrate these discoveries into a unified scaling law that accurately predicts the EL of an MoE architecture based on its configuration.
  arXiv  Detail & Related papers  (2025-07-23T17:10:23Z)
• A Systematic Analysis of Hybrid Linear Attention [11.722015123070957]
  Linear models often suffer from limited recall performance.<n>Our study highlights selective gating, hierarchical recurrence, and controlled forgetting as critical for effective hybrid models.<n>Our models are open-sourced at https://huggingface.co/collections/m-hugging-a-p/hybrid-linear-attention-research-686c488a63d609d2f2 0e2b1e.
  arXiv  Detail & Related papers  (2025-07-08T23:54:11Z)
• Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMs [51.21041884010009]
  Ring-lite is a Mixture-of-Experts (MoE)-based large language model optimized via reinforcement learning (RL)<n>Our approach matches the performance of state-of-the-art (SOTA) small-scale reasoning models on challenging benchmarks.
  arXiv  Detail & Related papers  (2025-06-17T17:12:34Z)
• Data-Driven Antenna Miniaturization: A Knowledge-Based System Integrating Quantum PSO and Predictive Machine Learning Models [0.0]
  This study integrates Quantum-Behaved Dynamic Particle Swarm Optimization with HFSS simulations to accelerate antenna design.<n>QDPSO algorithm autonomously optimized loop dimensions in 11.53 seconds, achieving a resonance frequency of 1.4208 GHz.<n>System enables precise specifications of performance targets with automated generation of fabrication-ready parameters.
  arXiv  Detail & Related papers  (2025-05-28T15:04:36Z)
• Hymba: A Hybrid-head Architecture for Small Language Models [65.94140459055244]
  Hymba is a family of small language models featuring a hybrid-head parallel architecture. We introduce learnable meta tokens that are prepended to prompts, storing critical information. This model is further optimized by incorporating cross-layer key-value sharing and partial sliding window attention.
  arXiv  Detail & Related papers  (2024-11-20T19:51:25Z)
• Towards Practical Lipreading with Distilled and Efficient Models [57.41253104365274]
  Lipreading has witnessed a lot of progress due to the resurgence of neural networks. Recent works have placed emphasis on aspects such as improving performance by finding the optimal architecture or improving generalization. There is still a significant gap between the current methodologies and the requirements for an effective deployment of lipreading in practical scenarios. We propose a series of innovations that significantly bridge that gap: first, we raise the state-of-the-art performance by a wide margin on LRW and LRW-1000 to 88.5% and 46.6%, respectively using self-distillation.
  arXiv  Detail & Related papers  (2020-07-13T16:56:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.