Yuanxin Liu

Awesome-Blender-Seedance-Workflow-Usecases. Curated Blender + Seedance workflows for AI filmmaking: previs, camera control, Blender MCP, reference video, and agent-guided use cases.

ppt-master. AI generates a real, editable PowerPoint from any document — native shapes & animations, editable charts & tables you can change the data on, speaker notes voiced as audio narration, and the option to follow your own .pptx template, not slide images · by Hugo He

deep-swe. Measuring frontier coding agents on original, long-horizon engineering tasks

open-design. 🎨 The open-source Claude Design alternative. 🖥️ Local-first desktop app. 🖼️ Your coding agent becomes the design engine: prototypes, landing pages, dashboards, slides, images & video — real files, HTML/PDF/PPTX/MP4 export. 🤖 Claude Code / Codex / Cursor / Gemini / OpenCode / Qwen & 20+ CLIs via BYOK.

awesome-design-md. A collection of DESIGN.md files analysis by popular brand design systems. Drop one into your project and let coding agents generate a matching UI.

goal-driven. A multi-agent system that keeps running for ~100 hours and solve a very complicated coding or math problem that can be verified

project-based-learning. Curated list of project-based tutorials

build-your-own-x. Master programming by recreating your favorite technologies from scratch.

LLMs-from-scratch. Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

BloodArena. JavaScript

Stockfish. A free and strong UCI chess engine

open_spiel. OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games.

game_arena. Python

flash-attention. Fast and memory-efficient exact attention

NL2RepoBench. Python

holi-spatial. [ICML 2026 Oral] Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence

claude-code-auto-video-edit. A Claude Code slash command that automates A-roll rough cuts | 达芬奇自动口播视频粗剪：Whisper 转写 → 智能分析剪辑点 → 生成达芬奇时间线 + SRT 字幕

claude-code-video-toolkit. AI-native video production toolkit for Claude Code

ai-paper-checking.

agentic-awesome-skills. Installable GitHub library of 1,900+ agentic skills for Claude Code, Cursor, Codex CLI, Autohand Code, Gemini CLI, Antigravity, and more. Includes specialized plugins, installer CLI, bundles, workflows, and official/community skill collections.

video-expert-analyzer. AI-powered video analysis skill based on Walter Murch's Six Rules | 基于 Walter Murch 剪辑六法则的 AI 视频分析工具

Auto-claude-code-research-in-sleep. ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works with Claude Code, Codex, OpenClaw, or any LLM agent.

learn-claude-code. Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1

openclaw. Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

claw-eval. Claw-Eval is an evaluation harness for evaluating LLM as agents. All tasks verified by humans.

awesome-openclaw-skills. The awesome collection of OpenClaw skills. 5,400+ skills filtered and categorized from the official OpenClaw Skills Registry.🦞

MLLM-Token-Compression. Towards Efficient Multimodal Large Language Models: A Survey on Token Compression

alphora. A Production-Ready Framework for Building Composable AI Agents

claude-code-router. Use Claude Code as the foundation for coding infrastructure, allowing you to decide how to interact with the model while enjoying updates from Anthropic.

PaperBanana. PaperBanana: Automating Academic Illustration For AI Scientists

AgiBot-World. [IROS 2025 Best Paper Award Finalist & IEEE TRO 2026] The Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems

DreamOmni2. This project is the official implementation of 'DreamOmni2: Multimodal Instruction-based Editing and Generation (CVPR2026 Highlight)''

Kimi-K2.5. Moonshot's most powerful model

MiMo-Audio. MiMo-Audio: Audio Language Models are Few-Shot Learners

MiMo-V2-Flash. MiMo-V2-Flash: Efficient Reasoning, Coding, and Agentic Foundation Model

VANS. [CVPR 2026] Video-as-Answer: Predict and Generate Next Video Event with Joint-GRPO

OneThinker. 🔥 OneThinker: All-in-one Reasoning Model for Image and Video [CVPR 2026]

Awesome-Multimodal-Spatial-Reasoning. This repository collects and organises state‑of‑the‑art papers on spatial reasoning for Multimodal Vision–Language Models (MVLMs).

cambrian-s. Cambrian-S: Towards Spatial Supersensing in Video

sam3. The repository provides code for running inference and finetuning with the Meta Segment Anything Model 3 (SAM 3), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.

Awesome-Video-Reasoning-Landscape. 🔥An open-source survey of the latest video reasoning tasks, paradigms, and benchmarks.

Awesome-Multimodal-LLM-for-Math-STEM. Paper collections of multi-modal LLM for Math/STEM/Code.

VideoSSR. [CVPR 2026] Official repo for "VideoSSR: Video Self-Supervised Reinforcement Learning"

Qwen3. Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud.

Wan2.2. Wan: Open and Advanced Large-Scale Video Generative Models

Emu3.5. Native Multimodal Models are World Learners

ms-swift. Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).

Open-o3-Video. [ICML 2026] Official implementation of "Open-o3 Video: Grounded Video Reasoning with Explicit Spatio-Temporal Evidence"

Conan. Multi-step reasoning MLLM

PixelRefer. The code for PixelRefer & VideoRefer

MotionLLM. [Arxiv-2024] MotionLLM: Understanding Human Behaviors from Human Motions and Videos

Awesome-Video-LMM-Post-Training. 🔥🔥🔥 Latest Papers, Codes and Datasets on Video-LMM Post-Training

SpatialVID. [CVPR 2026] SpatialVID: A Large-Scale Video Dataset with Spatial Annotations

DeepResearch. Tongyi Deep Research, the Leading Open-source Deep Research Agent

Awesome-Multimodal-Token-Compression. [TMLR 2026] Survey: https://arxiv.org/pdf/2507.20198

FETV. [NeurIPS 2023 Datasets and Benchmarks] "FETV: A Benchmark for Fine-Grained Evaluation of Open-Domain Text-to-Video Generation", Yuanxin Liu, Lei Li, Shuhuai Ren, Rundong Gao, Shicheng Li, Sishuo Chen, Xu Sun, Lu Hou

FETV-EVAL. Evaluation of generated videos on the FETV benchmark

Awesome_Think_With_Images. Resources and paper list for "Thinking with Images for LVLMs". This repository accompanies our survey on how LVLMs can leverage visual information for complex reasoning, planning, and generation.

DeepEyes. Python

Bagel. Open-source unified multimodal model

UI-TARS. Pioneering Automated GUI Interaction with Native Agents

Skywork-R1V. Skywork-R1V is an advanced multimodal AI model series developed by Skywork AI, specializing in vision-language reasoning.

ReCamMaster. [ICCV'25 Best Paper Finalist] ReCamMaster: Camera-Controlled Generative Rendering from A Single Video

GLM-V. GLM-4.6V/4.5V/4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

Keye. Python

VRBench. [ICCV 2025] A Benchmark for Multi-Step Reasoning in Long Narrative Videos

Kimi-K2. Kimi K2 is the large language model series developed by Moonshot AI team

vggt. [CVPR 2025 Best Paper Award] VGGT: Visual Geometry Grounded Transformer

SegAnyMo. [CVPR 2025] Code for Segment Any Motion in Videos

tarsier. Tarsier -- a family of large-scale video-language models, which is designed to generate high-quality video descriptions , together with good capability of general video understanding.

MotionBench. Official code for MotionBench (CVPR 2025)

CameraBench. [NeurIPS 2025 Spotlight] Towards Understanding Camera Motions in Any Video

Kimina-Prover-Preview. Technical report of Kimina-Prover Preview.

MiMo-VL. MiMo-VL

VCR-Bench. VCR-Bench: A Comprehensive Evaluation Framework for Video Chain-of-Thought Reasoning

Video-Holmes. [ECCV 2026] Video-Holmes: Can MLLM Think Like Holmes for Complex Video Reasoning?

md-badges. An extensive list of Shields.io badges.

VideoMMMU. Evaluating Knowledge Acquisition from Multi-Discipline Professional Videos

video_reason_bench. [ICLR 2026] "VideoReasonBench: Can MLLMs Perform Vision-Centric Complex Video Reasoning?", Yuanxin Liu, Kun Ouyang, Haoning Wu, Yi Liu, Lin Sui, Xinhao Li, Yan Zhong, Y. Charles, Xinyu Zhou, Xu Sun

DEER. This is the repository of DEER, a Dynamic Early Exit in Reasoning method for Large Reasoning Language Models.

TimeChat-Online. [ACM MM 2025] TimeChat-online: 80% Visual Tokens are Naturally Redundant in Streaming Videos

Seed1.5-VL. Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning, achieving state-of-the-art performance on 38 out of 60 public benchmarks.

my-tools. my commonly-used tools

perception_models. State-of-the-art Image & Video CLIP, Multimodal Large Language Models, and More!

LLM-PuzzleTest. This repository is maintained to release dataset and models for multimodal puzzle reasoning.

vllm. A high-throughput and memory-efficient inference and serving engine for LLMs

SpaceR. SpaceR: The first MLLM empowered by SG-RLVR for video spatial reasoning

Kimi-k1.5.

UVE. Python

Kimi-VL. Kimi-VL: Mixture-of-Experts Vision-Language Model for Multimodal Reasoning, Long-Context Understanding, and Strong Agent Capabilities

video-t3. [ICLR 2025] "Temporal Reasoning Transfer from Text to Video", Lei Li, Yuanxin Liu, Linli Yao, Peiyuan Zhang, Chenxin An, Lean Wang, Xu Sun, Lingpeng Kong, Qi Liu

Video-UTR. [ICLR2025] Official code implementation of Video-UTR: Unhackable Temporal Rewarding for Scalable Video MLLMs

VL_RewardBench. Python

Awesome-MLLM-Reasoning-Benchmarks. A Comprehensive Survey on Evaluating Reasoning Capabilities in Multimodal Large Language Models.

Awesome-MCoT. Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey

verl. verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework

VideoChat-Flash. [ICLR2026] VideoChat-Flash: Hierarchical Compression for Long-Context Video Modeling

MM-EUREKA. MM-EUREKA: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learning

NumPro. [CVPR2025] Number it: Temporal Grounding Videos like Flipping Manga

VisualThinker-R1-Zero. Explore the Multimodal “Aha Moment” on 2B Model

Visual-RFT. Official repository of 'Visual-RFT: Visual Reinforcement Fine-Tuning' & 'Visual-ARFT: Visual Agentic Reinforcement Fine-Tuning'’

Awesome-Multimodal-Reasoning. Latest Advances on (RL based) Multimodal Reasoning and Generation in Multimodal LLMs

theLMbook. This is the official repository for The Hundred-Page Language Models Book by Andriy Burkov

Awesome-System2-Reasoning-LLM. Latest Advances on System-2 Reasoning

MME-CoT. MME-CoT: Benchmarking Chain-of-Thought in LMMs for Reasoning Quality, Robustness, and Efficiency

lmm-r1. Extend OpenRLHF to support LMM RL training for reproduction of DeepSeek-R1 on multimodal tasks.

R1-Onevision. R1-onevision, a visual language model capable of deep CoT reasoning.

VLM-R1. Solve Visual Understanding with Reinforced VLMs

Open-R1-Video. ✨First Open-Source R1-like Video-LLM [2025/02/18]

Video-R1. Video-R1: Reinforcing Video Reasoning in MLLMs [🔥the first paper to explore R1 for video]

Awesome-RL-based-Reasoning-MLLMs. This repository provides valuable reference for researchers in the field of multimodality, please start your exploratory travel in RL-based Reasoning MLLMs!

open-r1-multimodal. A fork to add multimodal model training to open-r1

Awesome-LLM-Reasoning. From Chain-of-Thought prompting to OpenAI o1 and DeepSeek-R1 🍓

R1-V. Witness the aha moment of VLM with less than $3.

s1. s1: Simple test-time scaling

simpleRL-reason. Simple RL training for reasoning

DeepSeek-R1.

aesthetic-predictor. A linear estimator on top of clip to predict the aesthetic quality of pictures

cosmos. NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.

matryoshka-mm. Matryoshka Multimodal Models

Awesome-Multimodal-Next-Token-Prediction. [Survey] Next Token Prediction Towards Multimodal Intelligence: A Comprehensive Survey

Awesome-Token-Compress. A paper list of some recent works about Token Compress for Vit and VLM

genesis-world. Simulation platform for general-purpose robotics & embodied AI learning.

Visual-Question-Answering-for-Video-Quality-Assessment. [ACMMM2025] Official released code for VQA² series models

HunyuanVideo. HunyuanVideo: A Systematic Framework For Large Video Generation Model

VisionZip. Official repository for VisionZip (CVPR 2025)

FastV. [ECCV 2024 Oral] Code for paper: An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models

Vinoground. Python

Aria. Codebase for Aria - an Open Multimodal Native MoE

VLMEvalKit. Open-source evaluation toolkit of large multi-modality models (LMMs), support 220+ LMMs, 80+ benchmarks

LongVideoBench. [Neurips 24' D&B] Official Dataloader and Evaluation Scripts for LongVideoBench.

VideoGen-Eval. VideoGen-Eval: Agent-based System for Video Generation Evaluation

LongVU. [ICML 2025] Official PyTorch implementation of LongVU

DEVIL. Evaluation of Text-to-Video Generation Models: A Dynamics Perspective[NeurIPS 2024].

t2v_metrics. Evaluating text-to-image/video/3D models with VQAScore

DeCo. Code for DeCo: Decoupling token compression from semanchc abstraction in multimodal large language models

ETBench. 👾 E.T. Bench: Towards Open-Ended Event-Level Video-Language Understanding (NeurIPS 2024)

TOMATO. Python

GAIA. [NeurIPS'24 Spotlight] GAIA: Rethinking Action Quality Assessment for AI-Generated Videos

Awesome-Evaluation-of-Visual-Generation. A list of works on evaluation of visual generation models, including evaluation metrics, models, and systems

Video-XL. 🔥🔥First-ever hour scale video understanding models

VideoNIAH. VideoNIAH: A Flexible Synthetic Method for Benchmarking Video MLLMs

MovieGenBench. Movie Gen Bench - two media generation evaluation benchmarks released with Meta Movie Gen

T2V-CompBench. [CVPR 2025] T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation

VideoScore. official repo for "VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation" [EMNLP2024]

Qwen3-VL. Qwen3-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Cheetah. Python

MiniCPM-V. A Pocket-Sized MLLM for Ultra-Efficient Image and Video Understanding on Your Phone

PoSE. Positional Skip-wise Training for Efficient Context Window Extension of LLMs to Extremely Length (ICLR 2024)

nlp-tutorial. NLP新手入门教程

LongEmbed. LongEmbed: Extending Embedding Models for Long Context Retrieval (EMNLP 2024)

TempCompass. [ACL 2024 Findings] "TempCompass: Do Video LLMs Really Understand Videos?", Yuanxin Liu, Shicheng Li, Yi Liu, Yuxiang Wang, Shuhuai Ren, Lei Li, Sishuo Chen, Xu Sun, Lu Hou

VILA. VILA is a family of state-of-the-art vision language models (VLMs) for diverse multimodal AI tasks across the edge, data center, and cloud.

sam2. The repository provides code for running inference with the Meta Segment Anything Model 2 (SAM 2), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.

cambrian. Cambrian-1 is a family of multimodal LLMs with a vision-centric design.

BLINK_Benchmark. This repo contains evaluation code for the paper "BLINK: Multimodal Large Language Models Can See but Not Perceive". https://arxiv.org/abs/2404.12390 [ECCV 2024]

Video-STaR. [ICLR 2025] Video-STaR: Self-Training Enables Video Instruction Tuning with Any Supervision

Depth-Anything-V2. [NeurIPS 2024] Depth Anything V2. A More Capable Foundation Model for Monocular Depth Estimation

Elysium. [ECCV 2024] Elysium: Exploring Object-level Perception in Videos via MLLM

InternLM-XComposer. InternLM-XComposer2.5-OmniLive: A Comprehensive Multimodal System for Long-term Streaming Video and Audio Interactions

InternVL. [CVPR 2024 Oral] InternVL Family: A Pioneering Open-Source Alternative to GPT-4o. 接近GPT-4o表现的开源多模态对话模型

VideoLLaMA2. VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs

LongVA. Long Context Transfer from Language to Vision

vggsfm. VGGSfM: Visual Geometry Grounded Deep Structure From Motion

MMEvalPro. [NAACL 2025] Source code for MMEvalPro, a more trustworthy and efficient benchmark for evaluating LMMs

lmms-eval. One-for-All Multimodal Evaluation Toolkit Across Text, Image, Video, and Audio Tasks

MLVU. 🔥🔥MLVU: Multi-task Long Video Understanding Benchmark

Vript. Python

ShareGPT4Video. [NeurIPS 2024] An official implementation of "ShareGPT4Video: Improving Video Understanding and Generation with Better Captions"

PLLaVA. Official repository for the paper PLLaVA

Video-MME. ✨✨[CVPR 2025] Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis

LLaVA-NeXT. Python

FreeVA. FreeVA: Offline MLLM as Training-Free Video Assistant

Tabular-LLM. 本项目旨在收集开源的表格智能任务数据集（比如表格问答、表格-文本生成等），将原始数据整理为指令微调格式的数据并微调LLM，进而增强LLM对于表格数据的理解，最终构建出专门面向表格智能任务的大型语言模型。

Table-LLaVA. Dataset and Code for our ACL 2024 paper: "Multimodal Table Understanding". We propose the first large-scale Multimodal IFT and Pre-Train Dataset for table understanding and develop a generalist tabular MLLM named Table-LLaVA.

MotionCtrl. Official Code for MotionCtrl [SIGGRAPH 2024]

recognize-anything. Open-source and strong foundation image recognition models.

Track-Anything. Track-Anything is a flexible and interactive tool for video object tracking and segmentation, based on Segment Anything, XMem, and E2FGVI.

Grounded-Segment-Anything. Grounded SAM: Marrying Grounding DINO with Segment Anything & Stable Diffusion & Recognize Anything - Automatically Detect , Segment and Generate Anything

LISA. Project Page for "LISA: Reasoning Segmentation via Large Language Model"

co-tracker. CoTracker is a model for tracking any point (pixel) on a video.

acad-homepage.github.io. AcadHomepage: A Modern and Responsive Academic Personal Homepage

MyArxiv. Arxiv个性化定制化模版，实现对特定领域的相关内容、作者与学术会议的有效跟进。

GroundingDINO. [ECCV 2024] Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"

GRiT. GRiT: A Generative Region-to-text Transformer for Object Understanding (ECCV2024)

shikra. Python

Awesome-LLMs-for-Video-Understanding. 🔥🔥🔥 [IEEE TCSVT] Latest Papers, Codes and Datasets on Vid-LLMs.

VLFeedback. Python

EMO. Emote Portrait Alive: Generating Expressive Portrait Videos with Audio2Video Diffusion Model under Weak Conditions

VBench. [CVPR2024 Highlight] VBench - We Evaluate Video Generation

MMVP. Python

TimeChat. [CVPR 2024] TimeChat: A Time-sensitive Multimodal Large Language Model for Long Video Understanding

LLaMA2-Accessory. An Open-source Toolkit for LLM Development

LRV-Instruction. [ICLR'24] Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning

ICL_PaperList. Paper List for In-context Learning 🌷

generative-models. Generative Models by Stability AI

Video-Bench. A Comprehensive Benchmark and Toolkit for Evaluating Video-based Large Language Models!

Video-LLaVA. 【EMNLP 2024🔥】Video-LLaVA: Learning United Visual Representation by Alignment Before Projection

Woodpecker. ✨✨Woodpecker: Hallucination Correction for Multimodal Large Language Models

academicpages.github.io. Github Pages template based upon HTML and Markdown for personal, portfolio-based websites.