arXiv subjects

Yangyang Hu

Publications and source records attributed to Yangyang Hu.

10 recordsLinked to original sources

arXiv2026

Kimi K3: Open Frontier Intelligence

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence.

cs.CL↗

arXiv2026

Kimi K2.5: Visual Agentic Intelligence

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

cs.CL↗

arXiv2025

Kimi K2: Open Agentic Intelligence

We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated parameters and 1 trillion total parameters. We propose the MuonClip optimizer, which improves upon Muon with a novel QK-clip technique to address training instability while enjoying the advanced token efficiency of Muon. Based on MuonClip, K2 was pre-trained on 15.5 trillion tokens with zero loss spike. During post-training, K2 undergoes a multi-stage post-training process, highlighted by a large-scale agentic data synthesis pipeline and a joint reinforcement learning (RL) stage, where the model improves its capabilities through interactions with real and synthetic environments. Kimi K2 achieves state-of-the-art performance among open-source non-thinking models, with strengths in agentic capabilities. Notably, K2 obtains 66.1 on Tau2-Bench, 76.5 on ACEBench (En), 65.8 on SWE-Bench Verified, and 47.3 on SWE-Bench Multilingual -- surpassing most open and closed-sourced baselines in non-thinking settings. It also exhibits strong capabilities in coding, mathematics, and reasoning tasks, with a score of 53.7 on LiveCodeBench v6, 49.5 on AIME 2025, 75.1 on GPQA-Diamond, and 27.1 on OJBench, all without extended thinking. These results position Kimi K2 as one of the most capable open-source large language models to date, particularly in software engineering and agentic tasks. We release our base and post-trained model checkpoints to facilitate future research and applications of agentic intelligence.

cs.LG↗

arXiv2025

DIFFUMA: High-Fidelity Spatio-Temporal Video Prediction via Dual-Path Mamba and Diffusion Enhancement

Spatio-temporal video prediction plays a pivotal role in critical domains, ranging from weather forecasting to industrial automation. However, in high-precision industrial scenarios such as semiconductor manufacturing, the absence of specialized benchmark datasets severely hampers research on modeling and predicting complex processes. To address this challenge, we make a twofold contribution.First, we construct and release the Chip Dicing Lane Dataset (CHDL), the first public temporal image dataset dedicated to the semiconductor wafer dicing process. Captured via an industrial-grade vision system, CHDL provides a much-needed and challenging benchmark for high-fidelity process modeling, defect detection, and digital twin development.Second, we propose DIFFUMA, an innovative dual-path prediction architecture specifically designed for such fine-grained dynamics. The model captures global long-range temporal context through a parallel Mamba module, while simultaneously leveraging a diffusion module, guided by temporal features, to restore and enhance fine-grained spatial details, effectively combating feature degradation. Experiments demonstrate that on our CHDL benchmark, DIFFUMA significantly outperforms existing methods, reducing the Mean Squared Error (MSE) by 39% and improving the Structural Similarity (SSIM) from 0.926 to a near-perfect 0.988. This superior performance also generalizes to natural phenomena datasets. Our work not only delivers a new state-of-the-art (SOTA) model but, more importantly, provides the community with an invaluable data resource to drive future research in industrial AI.

cs.CV↗

arXiv2025

Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning

We introduce Kimina-Prover Preview, a large language model that pioneers a novel reasoning-driven exploration paradigm for formal theorem proving, as showcased in this preview release. Trained with a large-scale reinforcement learning pipeline from Qwen2.5-72B, Kimina-Prover demonstrates strong performance in Lean 4 proof generation by employing a structured reasoning pattern we term \textit{formal reasoning pattern}. This approach allows the model to emulate human problem-solving strategies in Lean, iteratively generating and refining proof steps. Kimina-Prover sets a new state-of-the-art on the miniF2F benchmark, reaching 80.7% with pass@8192. Beyond improved benchmark performance, our work yields several key insights: (1) Kimina-Prover exhibits high sample efficiency, delivering strong results even with minimal sampling (pass@1) and scaling effectively with computational budget, stemming from its unique reasoning pattern and RL training; (2) we demonstrate clear performance scaling with model size, a trend previously unobserved for neural theorem provers in formal mathematics; (3) the learned reasoning style, distinct from traditional search algorithms, shows potential to bridge the gap between formal verification and informal mathematical intuition. We open source distilled versions with 1.5B and 7B parameters of Kimina-Prover

cs.AI↗

arXiv2025

Kimi-VL Technical Report

We present Kimi-VL, an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities - all while activating only 2.8B parameters in its language decoder (Kimi-VL-A3B). Kimi-VL demonstrates strong performance across challenging domains: as a general-purpose VLM, Kimi-VL excels in multi-turn agent tasks (e.g., OSWorld), matching flagship models. Furthermore, it exhibits remarkable capabilities across diverse challenging vision language tasks, including college-level image and video comprehension, OCR, mathematical reasoning, and multi-image understanding. In comparative evaluations, it effectively competes with cutting-edge efficient VLMs such as GPT-4o-mini, Qwen2.5-VL-7B, and Gemma-3-12B-IT, while surpassing GPT-4o in several key domains. Kimi-VL also advances in processing long contexts and perceiving clearly. With a 128K extended context window, Kimi-VL can process diverse long inputs, achieving impressive scores of 64.5 on LongVideoBench and 35.1 on MMLongBench-Doc. Its native-resolution vision encoder, MoonViT, further allows it to see and understand ultra-high-resolution visual inputs, achieving 83.2 on InfoVQA and 34.5 on ScreenSpot-Pro, while maintaining lower computational cost for common tasks. Building upon Kimi-VL, we introduce an advanced long-thinking variant: Kimi-VL-Thinking-2506. Developed through long chain-of-thought (CoT) supervised fine-tuning (SFT) and reinforcement learning (RL), the latest model exhibits strong long-horizon reasoning capabilities (64.0 on MMMU, 46.3 on MMMU-Pro, 56.9 on MathVision, 80.1 on MathVista, 65.2 on VideoMMMU) while obtaining robust general abilities. Code and models are publicly accessible at https://github.com/MoonshotAI/Kimi-VL.

cs.CV↗

arXiv2025

A Novel Thermal Network Model and Electro-Thermal Coupling Study for NSFETs and CFETs Considering Thermal Crosstalk

As the technology node continues to shrink, nanosheet field effect transistors (NSFETs) and complementary FETs (CFETs) become valid candidates for the 3nm and sub-nanometre nodes. However, due to the shrinking device size, self-heating and inter-device thermal crosstalk of NSFETs and CFETs become more severe. It is important to accurately calculate the self-heating and thermal crosstalk of devices and to study the electrical and thermal characteristics of logic gates, etc. In this work, a thermal network model considering the thermal crosstalk of neighboring devices is proposed, which can accurately calculate the self-heating and thermal crosstalk. The electrical and thermal characteristics of NSFETs and CFETs are compared, and it is found that CFETs have more severe self-heating and thermal crosstalk. The electro-thermal characteristics of inverters, logic gates and ring oscillators composed of NSFETs and CFETs are further investigated. Compared with NSFETs, logic gates and ring oscillators composed of CFETs are more seriously affected by self-heating and should be given extra attention. The thermal network model proposed in this paper can be further used to study the thermal optimization strategy of devices and circuits to enhance the electrical performance, achieving the design technology co-optimizations (DTCO).

cs.ET↗

arXiv2025

A Novel Design for SRAM Bitcell with 3-Complementary-FETs

The complementary field-effect transistors (CFETs), featuring vertically stacked n/p-FETs, enhance integration density and significantly reduce the area of standard cells such as static random-access memory (SRAM). However, the advantage of area scaling through CFETs is hindered by the imbalance in N/P transistor counts (typically 4N/2P) within SRAM cells. In this work, we propose a novel 6T-SRAM design using three sets of CFETs, achieved by vertically stacking two n-FET pass-gate (PG) transistors via the CFET architecture. Through TCAD simulations, we optimize channel doping concentration and the number of top/bottom nanosheets (NS), demonstrating that junctionless accumulation mode (JAM) devices outperform inversion mode (IM) devices for PG and pull-down (PD) transistors. The proposed design achieves a 37% area reduction in SRAM standard cell layout compared to conventional CFET-based SRAM. With optimized parameters (n-type doping of $1\times10^{15}$ cm$^{-3}$ and '1B4T' NS configuration), the 3-CFET SRAM exhibits superior write margin (349.60 mV) and write delay (54.4 ps). This work advances SRAM design within the CFET framework, offering a scalable solution for next-generation memory technologies.

physics.app-ph↗

arXiv2025

Kimi k1.5: Scaling Reinforcement Learning with LLMs

Language model pretraining with next token prediction has proved effective for scaling compute but is limited to the amount of available training data. Scaling reinforcement learning (RL) unlocks a new axis for the continued improvement of artificial intelligence, with the promise that large language models (LLMs) can scale their training data by learning to explore with rewards. However, prior published work has not produced competitive results. In light of this, we report on the training practice of Kimi k1.5, our latest multi-modal LLM trained with RL, including its RL training techniques, multi-modal data recipes, and infrastructure optimization. Long context scaling and improved policy optimization methods are key ingredients of our approach, which establishes a simplistic, effective RL framework without relying on more complex techniques such as Monte Carlo tree search, value functions, and process reward models. Notably, our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities -- e.g., 77.5 on AIME, 96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching OpenAI's o1. Moreover, we present effective long2short methods that use long-CoT techniques to improve short-CoT models, yielding state-of-the-art short-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on LiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and Claude Sonnet 3.5 by a large margin (up to +550%).

cs.AI↗

arXiv2022

Enhancing Neural Mathematical Reasoning by Abductive Combination with Symbolic Library

Mathematical reasoning recently has been shown as a hard challenge for neural systems. Abilities including expression translation, logical reasoning, and mathematics knowledge acquiring appear to be essential to overcome the challenge. This paper demonstrates that some abilities can be achieved through abductive combination with discrete systems that have been programmed with human knowledge. On a mathematical reasoning dataset, we adopt the recently proposed abductive learning framework, and propose the ABL-Sym algorithm that combines the Transformer neural models with a symbolic mathematics library. ABL-Sym shows 9.73% accuracy improvement on the interpolation tasks and 47.22% accuracy improvement on the extrapolation tasks, over the state-of-the-art approaches. Online demonstration: http://math.polixir.ai

cs.LG↗