Author

Kuaishou Tech

Official Kuaishou tech account, providing real-time updates on the latest Kuaishou technology practices.

244

Articles

Likes

1.5k

Views

Comments

Latest from Kuaishou Tech

100 recent articles max

Kuaishou Tech

Dec 3, 2025 · Artificial Intelligence

Can Diffusion Models Be Their Own Reward Model? Latent Reward Modeling & Step-Level Preference Optimization

This article presents a novel paradigm—Latent Reward Model (LRM) and Latent Preference Optimization (LPO)—that repurposes diffusion models as noise‑aware latent reward models for step‑level preference optimization, addressing the shortcomings of pixel‑level reward models, introducing multi‑preference consistent filtering, and demonstrating significant performance and efficiency gains on benchmarks such as PickScore and T2I‑CompBench++.

AI alignmentdiffusion modelsimage generation

0 likes · 9 min read

Can Diffusion Models Be Their Own Reward Model? Latent Reward Modeling & Step-Level Preference Optimization

Kuaishou Tech

Nov 28, 2025 · Artificial Intelligence

Keye-VL-671B-A37B Leads Vision, Video, and Math Benchmarks

Kwai has open‑sourced its new flagship multimodal model Keye‑VL‑671B‑A37B, which upgrades visual perception, cross‑modal alignment and complex reasoning, achieving top scores on image, video, and mathematical reasoning benchmarks while detailing its architecture, three‑stage pre‑training, post‑training strategies, and future multimodal agent plans.

Deep LearningLarge Language ModelVideo Understanding

0 likes · 10 min read

Keye-VL-671B-A37B Leads Vision, Video, and Math Benchmarks

Kuaishou Tech

Nov 25, 2025 · Artificial Intelligence

How Flow‑GRPO Boosts Image Generation Accuracy to 95% with Online Reinforcement Learning

Flow‑GRPO introduces online reinforcement learning into flow‑matching models by converting deterministic ODE sampling to stochastic SDE sampling and reducing denoising steps, raising SD‑3.5‑Medium's GenEval accuracy from 63% to 95%—surpassing GPT‑4o—and demonstrating strong gains in complex composition, text rendering, and human‑preference alignment across multiple generative tasks.

AI researchDeep LearningFlow Matching

0 likes · 8 min read

How Flow‑GRPO Boosts Image Generation Accuracy to 95% with Online Reinforcement Learning

Kuaishou Tech

Nov 24, 2025 · Artificial Intelligence

How Human Feedback Supercharges Video Generation – The VideoAlign Pipeline Explained

This article details a new research pipeline that leverages large‑scale human preference data, a multi‑dimensional video reward model, and specialized alignment algorithms to dramatically improve video generation quality, motion fidelity, and text‑video consistency, with open‑source code and benchmarks for reproducibility.

AI alignmentHuman FeedbackRLHF

0 likes · 10 min read

How Human Feedback Supercharges Video Generation – The VideoAlign Pipeline Explained

Kuaishou Tech

Nov 20, 2025 · Artificial Intelligence

How UniDex and UniSearch Redefine Video Search with Semantic Indexing and Generative Models

This article explains how Kuaishou’s UniDex replaces traditional term‑based inverted indexes with model‑driven semantic posting lists and how the end‑to‑end UniSearch framework generates video IDs directly from queries, delivering higher relevance, lower latency, and significant online performance gains.

AIGenerative Modelslarge language models

0 likes · 17 min read

How UniDex and UniSearch Redefine Video Search with Semantic Indexing and Generative Models

Kuaishou Tech

Nov 19, 2025 · Artificial Intelligence

Can a Single Number Create a Whole New Visual Style? Inside CoTyle’s Code‑to‑Style Generation

CoTyle introduces a novel open‑source framework that generates unique image styles from a numeric style code, eliminating the need for reference images, lengthy prompts, or LoRA modules, and demonstrates superior style consistency compared to existing solutions like Midjourney.

Transformerdiffusion modelgenerative AI

0 likes · 8 min read

Can a Single Number Create a Whole New Visual Style? Inside CoTyle’s Code‑to‑Style Generation

Kuaishou Tech

Nov 14, 2025 · Artificial Intelligence

How GRPO‑Guard Stops Over‑Optimization in Flow‑Based Visual Generators

This article explains the over‑optimization problem in GRPO‑based flow models, analyzes why importance‑ratio clipping fails, and introduces GRPO‑Guard with RatioNorm and cross‑step gradient balancing, showing through extensive experiments that it stabilizes training and improves image quality across multiple diffusion backbones and tasks.

Flow MatchingGRPO-Guardgenerative AI

0 likes · 9 min read

How GRPO‑Guard Stops Over‑Optimization in Flow‑Based Visual Generators

Kuaishou Tech

Nov 13, 2025 · Artificial Intelligence

Unlocking Unusual Concept Combinations in Generative AI with IMBA Loss

The paper identifies imbalanced concept distributions as the main obstacle to arbitrary concept‑combination in text‑to‑image/video generation, proposes the token‑level IMBA Distance and a lightweight IMBA Loss that adaptively re‑weights training tokens, and demonstrates through extensive experiments and a new Inert‑CompBench benchmark that this loss dramatically improves compositional ability without extra data.

IMBA Lossbenchmarkconcept combination

0 likes · 9 min read

Unlocking Unusual Concept Combinations in Generative AI with IMBA Loss

Kuaishou Tech

Nov 12, 2025 · Artificial Intelligence

How KaiFG Lets Python Feature Engineering Run at C++ Speed

KaiFG, Kuaishou's self‑built AI Feature Generator, unifies fragmented feature extraction frameworks, replaces slow C++ compilation cycles with Python‑level development, and achieves near‑C++ performance through Codon‑based compilation, reference‑counted memory management, and aggressive LLVM optimizations, dramatically shortening iteration time.

AI Infrastructurefeature engineeringhigh-performance computing

0 likes · 14 min read

How KaiFG Lets Python Feature Engineering Run at C++ Speed

Kuaishou Tech

Nov 5, 2025 · Artificial Intelligence

How HiPO Gives LLMs a Smart Thinking Switch to Cut Costs and Boost Accuracy

This article explains the overthinking problem of large language models, introduces the HiPO framework with hybrid data cold‑start and reinforcement‑learning reward mechanisms that let models decide when to think deeply or answer directly, and shows experimental results demonstrating significant efficiency gains and accuracy improvements across multiple benchmarks.

EfficiencyHybrid Policy OptimizationLLM

0 likes · 13 min read

How HiPO Gives LLMs a Smart Thinking Switch to Cut Costs and Boost Accuracy