[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-77210":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":15,"starSnapshotCount":15,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},77210,"Lance","bytedance\u002FLance","bytedance","A 3B-active-parameter native unified multimodal model for image and video understanding, generation, and editing.","https:\u002F\u002Flance-project.github.io",null,"Python",1186,79,14,0,9,57,1103,47,18.71,"Apache License 2.0",false,"main",[],"2026-06-12 02:03:42","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"assets\u002Flogo\u002Flance-logo.webp\" alt=\"Lance logo\" width=\"300\">\n\n  \u003Ch1 align=\"center\">\u003Csup>Lance: Unified Multimodal Modeling by Multi-Task Synergy\u003C\u002Fsup>\u003C\u002Fh1>\n  \u003Cp>\n    \u003Cstrong>\n    \u003Ca href=\"https:\u002F\u002Fscholar.google.com.hk\u002Fcitations?user=FXxoQlsAAAAJ&hl=zh-CN&oi=ao\" style=\"text-decoration: none; color: inherit;\">Fengyi Fu\u003C\u002Fa>\u003Csup>*\u003C\u002Fsup>, \n    \u003Ca href=\"https:\u002F\u002Fcorleone-huang.github.io\u002F\" style=\"text-decoration: none; color: inherit;\">Mengqi Huang\u003C\u002Fa>\u003Csup>*,✉\u003C\u002Fsup>, \n    \u003Ca href=\"https:\u002F\u002Fscholar.google.com.hk\u002Fcitations?user=9ER6nVkAAAAJ&hl=zh-CN&oi=ao\" style=\"text-decoration: none; color: inherit;\">Shaojin Wu\u003C\u002Fa>\u003Csup>*\u003C\u002Fsup>, \n    Yunsheng Jiang\u003Csup>*\u003C\u002Fsup>, \n    Yufei Huo, \n    \u003Ca href=\"https:\u002F\u002Fguojianzhu.com\u002F\" style=\"text-decoration: none; color: inherit;\">Jianzhu Guo\u003C\u002Fa>\u003Csup>✉,§\u003C\u002Fsup>\n    \u003C\u002Fstrong>\u003Cbr>\n    Hao Li, \n    Yinghang Song, \n    Fei Ding, \n    Qian He, \n    Zheren Fu, \n    Zhendong Mao, \n    Yongdong Zhang\n    \u003Cbr>\n    \u003Cem>ByteDance\u003C\u002Fem>\n    \u003Cbr>\n    \u003Csup>*\u003C\u002Fsup> Equal contribution    \u003Csup>✉\u003C\u002Fsup> Corresponding authors    \u003Csup>§\u003C\u002Fsup> Project lead\n  \u003C\u002Fp>\n  \u003Cp>\n    \u003Ca href=\"https:\u002F\u002Flance-project.github.io\u002F\" style=\"text-decoration: none; margin: 0 8px;\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FHomepage-Lance-blue?style=flat\" alt=\"Homepage\">\u003C\u002Fa>\n    \u003Ca href=\"http:\u002F\u002Farxiv.org\u002Fabs\u002F2605.18678\" style=\"text-decoration: none; margin: 0 8px;\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPaper-arXiv-red?style=flat&logo=arxiv\" alt=\"arXiv\">\u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fbytedance-research\u002FLance\" style=\"text-decoration: none; margin: 0 8px;\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FModel-HuggingFace-yellow?style=flat&logo=huggingface\" alt=\"Model\">\u003C\u002Fa>\n    \u003Cbr>\n    English | \u003Ca href=\".\u002FREADME_zh.md\">\u003Cins>简体中文\u003C\u002Fins>\u003C\u002Fa>\n  \u003C\u002Fp>\n\u003C\u002Fdiv>\n\n## 🌟 Highlights\n\n**Lance** is a 3B native unified multimodal model that supports **image and video understanding, generation, and editing** within a single framework.\n\n- **Efficient at 3B scale.** With only **3B active parameters**, Lance delivers strong performance across image generation, image editing, and video generation benchmarks.\n- **Trained from scratch.** Lance is built with a staged multi-task recipe and trained entirely from scratch within a **128-A100-GPU** budget.\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"assets\u002Fbenchmarks\u002Fbenchmark-overview.png\" alt=\"Lance benchmark overview across image generation, image editing, video generation, and video understanding\" width=\"980\">\n\u003C\u002Fdiv>\n\n## 🎨 Demo\n\n### Text-to-Video\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-01.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-01.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-02.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-02.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-03.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-03.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-04.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-04.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-05.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-05.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-06.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-06.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-07.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-07.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Ftext-to-video\u002Fvideos\u002Ftext-to-video-demo-08.mp4\">\u003Cimg src=\"assets\u002Ftext-to-video\u002Fpreviews\u002Ftext-to-video-demo-08.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n### Video Editing\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-01.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-01.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-02.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-02.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-03.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-03.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-04.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-04.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-05.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-05.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-06.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-06.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-07.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-07.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fvideo-editing\u002Fvideos\u002Fvideo-editing-demo-08.mp4\">\u003Cimg src=\"assets\u002Fvideo-editing\u002Fpreviews\u002Fvideo-editing-demo-08.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n### Multi-turn Consistency Editing\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"assets\u002Fmulti-turn-editing\u002Fvideos\u002Fmulti-turn-editing-demo-01.mp4\">\n    \u003Cimg src=\"assets\u002Fmulti-turn-editing\u002Fpreviews\u002Fmulti-turn-editing-demo-01.gif\" width=\"100%\">\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n### Intelligent Video Generation\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd>\u003Ca href=\"assets\u002Fintelligent-video\u002Fvideos\u002Fintelligent-video-demo-01.mp4\">\u003Cimg src=\"assets\u002Fintelligent-video\u002Fpreviews\u002Fintelligent-video-demo-01.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fintelligent-video\u002Fvideos\u002Fintelligent-video-demo-02.mp4\">\u003Cimg src=\"assets\u002Fintelligent-video\u002Fpreviews\u002Fintelligent-video-demo-02.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fintelligent-video\u002Fvideos\u002Fintelligent-video-demo-03.mp4\">\u003Cimg src=\"assets\u002Fintelligent-video\u002Fpreviews\u002Fintelligent-video-demo-03.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd>\u003Ca href=\"assets\u002Fintelligent-video\u002Fvideos\u002Fintelligent-video-demo-04.mp4\">\u003Cimg src=\"assets\u002Fintelligent-video\u002Fpreviews\u002Fintelligent-video-demo-04.gif\" width=\"100%\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n### Video Understanding\n\n\u003Cdiv align=\"center\">\n  \u003Ctable align=\"center\">\n    \u003Ctr>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-vqa-01.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-vqa-01.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> How many times did the person launch objects on the table? Options: (A) 3 (B) 2 (C) 4\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> (A) 3\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-vqa-02.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-vqa-02.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> The person makes sets of repeated actions. How many distinct repeated actions did the person do? Options: (A) 2 (B) 3 (C) 4\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> (A) 2\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-vqa-03.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-vqa-03.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> In which direction does the purple sphere move in the video? Options: (A) Down and to the right. (B) Up and to the left. (C) Up and to the right. (D) The object is stationary.\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> (A) Down and to the right.\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-vqa-04.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-vqa-04.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> What is the unrealistic phenomenon displayed in the video? Options: (A) The man can manipulate time via phone. (B) Man grabs an object through a phone screen. (C) Chocolate transforms into different objects. (D) Visible means of propulsion enables flight.\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> (B) Man grabs an object through a phone screen.\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-caption-short-01.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-caption-short-01.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> Offer a succinct account of the culinary process shown in this video.\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> Add tomato puree and mix it well with chicken pieces.\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Ca href=\"assets\u002Fvideo-understanding\u002Fvideos\u002Fvideo-understanding-caption-long-01.mp4\">\n          \u003Cimg src=\"assets\u002Fvideo-understanding\u002Fpreviews\u002Fvideo-understanding-caption-long-01.gif\" width=\"100%\">\n        \u003C\u002Fa>\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> Provide a detailed description of the given video, capturing its key moments.\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> In a sunlit meadow, a small tortoiseshell butterfly rests on a purple flower. A bee, with black and yellow stripes, lands on the same flower. The butterfly flaps its wings gently, while the bee busies itself, collecting nectar. The flower sways slightly in the breeze. The butterfly then takes off, and the bee follows, both heading to the next flower. The scene is a vivid display of insect interaction in a natural setting, with the colors of the butterfly and the bee contrasting against the green background of the meadow. The video captures this peaceful moment in a short 6-second duration.\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n### Text-to-Image Generation\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"assets\u002Ftext-to-image\u002Ftext-to-image-overview.webp\" alt=\"Lance text-to-image generation examples\" width=\"100%\">\n\u003C\u002Fdiv>\n\n### Image Editing\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"assets\u002Fimage-editing\u002Fimage-editing-overview.webp\" alt=\"Lance image editing examples\" width=\"100%\">\n\u003C\u002Fdiv>\n\n### Image Understanding\n\n\u003Cdiv align=\"center\">\n  \u003Ctable align=\"center\">\n    \u003Ctr>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-01.png\" alt=\"Pie chart reasoning case\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> Is the largest segment greater than sum of all the other segments?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> Yes, the largest segment in the pie chart is the blue one, which is the most populous segment.\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-02.png\" alt=\"Survey chart about border security priorities\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> What percentage of respondents want better border security?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> 29%\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-03.png\" alt=\"License plate recognition case\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> What is the license plate number of the car?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> The license plate number of the car is BX62 BFY.\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-04.png\" alt=\"Market research chart reasoning case\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> According to the data from the proprietary market research, how much amount was spent on the promotional meetings and events during 1998?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> According to the data from the proprietary market research, the total amount spent on the promotional meetings and events during 1998 was approximately $1.3 billion.\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-05.png\" alt=\"Colosseum description case\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> What is the appearance of the Colosseum in Rome, Italy?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> The Colosseum in Rome, Italy is a magnificent ancient amphitheater made of stone and concrete. It is characterized by its large, arched openings and its semi-circular shape. The Colosseum has been partially restored, and its original color has faded, but the architectural details, like the columns and the semi-circular arches, are still visible. The lighting in the image highlights the grandeur and historical significance of the Colosseum, making it stand out against the backdrop of the evening sky.\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"left\" valign=\"top\" width=\"33%\">\n        \u003Cimg src=\"assets\u002Fimage-understanding\u002Fcases\u002Fimage-understanding-case-06.png\" alt=\"Total solar eclipse case\" width=\"100%\">\n        \u003Cp>\u003Cstrong>Question:\u003C\u002Fstrong> How does a total solar eclipse look like from Earth?\u003C\u002Fp>\n        \u003Cp>\u003Cstrong>Response:\u003C\u002Fstrong> A total solar eclipse looks like a dark circle with a bright, white outer edge. This is a result of the Earth's rotation and the Earth's shadow, which partially obscures the sun's light. The bright white edge of the eclipse is caused by the sun's high energy and its interaction with the Earth's atmosphere, while the dark part of the eclipse is due to the Earth's shadow and the surrounding air currents. The solar eclipse's shape, with its bright white edge and dark center, is similar to the shape of a full moon or a dark disk. It is a natural phenomenon that occurs in the atmosphere of the Earth and is an important part of the solar system.\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n## 🚀 Installation\n\n### Recommended Environment\n\n- **Software:** Python 3.10+, CUDA 12.4+ (required)\n- **Hardware:** A GPU with at least 40GB VRAM is required for inference\n\n### Installation Steps\n```bash\nbash .\u002Fsetup_env.sh\n```\n\n### Download Model Weights\n\nPlease download all necessary model checkpoints from [Lance-3B on Hugging Face](https:\u002F\u002Fhuggingface.co\u002Fbytedance-research\u002FLance) and place them in the `downloads\u002F` directory.\n\n## 📚 Usage\n\n\n### Inference\n\nWe provide a unified command-line interface for all generation \u002F editing \u002F understanding tasks:\n\n#### Option 1: Configure and Run the Unified Script\n\n```bash\nbash inference_lance.sh\n```\n\n- Before running, please configure the inference parameters at the top of `inference_lance.sh`.\n- **Supported tasks:** `t2i`, `t2v`, `image_edit`, `video_edit`, `x2t_image`, and `x2t_video`. You can modify `TASK_DEFAULT_CONFIGS` in `inference_lance.py` to customize the default data samples for each task.\n- **Note:** For all tasks, we recommend following the `prompt` format used in the provided examples when writing input prompts, as this typically leads to better generation quality.\n\n#### Option 2: Configure and Run the Unified Script\n\nWe provide task-specific one-click commands for different generation, editing, and understanding tasks.\n\n##### Text-to-Video Generation\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME t2v \\\n  --MODEL_PATH downloads\u002FLance_3B_Video \\\n  --RESOLUTION video_480p \\\n  --NUM_FRAMES 121 \\\n  --VIDEO_HEIGHT 480 \\\n  --VIDEO_WIDTH 848 \\\n  --SAVE_PATH_GEN results\u002Ft2v_121f\n```\n\n##### Text-to-Image Generation\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME t2i \\\n  --MODEL_PATH downloads\u002FLance_3B \\\n  --RESOLUTION image_768res \\\n  --VIDEO_HEIGHT 768 \\\n  --VIDEO_WIDTH 768 \\\n  --SAVE_PATH_GEN results\u002Ft2i\n```\n\n##### Video Editing\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME video_edit \\\n  --MODEL_PATH downloads\u002FLance_3B_Video \\\n  --RESOLUTION video_480p \\\n  --SAVE_PATH_GEN results\u002Fvideo_edit\n```\n\n##### Image Editing\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME image_edit \\\n  --MODEL_PATH downloads\u002FLance_3B \\\n  --RESOLUTION image_768res \\\n  --SAVE_PATH_GEN results\u002Fimage_edit\n```\n\n##### Video Understanding\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME x2t_video \\\n  --MODEL_PATH downloads\u002FLance_3B_Video \\\n  --RESOLUTION video_480p \\\n  --NUM_FRAMES 50 \\\n  --SAVE_PATH_GEN results\u002Fx2t_video\n```\n\n##### Image Understanding\n\n```bash\nbash inference_lance.sh \\\n  --TASK_NAME x2t_image \\\n  --MODEL_PATH downloads\u002FLance_3B \\\n  --RESOLUTION image_768res \\\n  --SAVE_PATH_GEN results\u002Fx2t_image\n```\n\n#### Available Tasks\n\n| Task Name              | Description                                      | Example JSON                                 |\n|------------------------|--------------------------------------------------|----------------------------------------------|\n| `t2v`                  | Text-to-Video generation                         | `config\u002Fexamples\u002Ft2v_example.json`           |\n| `t2i`                  | Text-to-Image generation                         | `config\u002Fexamples\u002Ft2i_example.json`           |\n| `image_edit`           | Image editing                                    | `config\u002Fexamples\u002Fimage_edit_example.json`    |\n| `video_edit`           | Video editing                                    | `config\u002Fexamples\u002Fvideo_edit_example.json`    |\n| `x2t_image`            | Image understanding            | `config\u002Fexamples\u002Fx2t_image_example.json`    |\n| `x2t_video`            | Video understanding            | `config\u002Fexamples\u002Fx2t_video_example.json`    |\n\nFor understanding examples:\n\n- `config\u002Fexamples\u002Fx2t_image_example.json`: image understanding examples for visual question answering and image-based reasoning.\n- `config\u002Fexamples\u002Fx2t_video_example.json`: video understanding examples for video question answering and video captioning.\n\n#### Parameters\n\nYou can configure the following hyperparameters at the top of the `inference_lance.sh` script:\n\n| Parameter | Default Value | Description |\n| --- | --- | --- |\n| `MODEL_PATH` | `\"downloads\u002FLance_3B\"` | Path to the downloaded Lance model weights  (`Lance_3B` or `Lance_3B_Video`). |\n| `NUM_GPUS` | `1` | Number of GPUs to use for inference. |\n| `VALIDATION_NUM_TIMESTEPS` | `30` | Number of denoising steps (e.g., 30 or 50). |\n| `VALIDATION_TIMESTEP_SHIFT` | `3.5` | Timestep shift parameter for flow matching scheduling. |\n| `CFG_TEXT_SCALE` | `4.0` | Classifier-Free Guidance (CFG) scale for text conditioning. |\n| `VALIDATION_DATA_SEED` | `42` | Random seed for generation reproducibility. |\n| `NUM_FRAMES` | `50` | Number of frames for video generation (Max: 121). *Unused for image tasks.* |\n| `VIDEO_HEIGHT` \u002F `VIDEO_WIDTH`| `768` | Spatial resolution. *Unused for editing tasks (determined by input image\u002Fvideo).* |\n| `RESOLUTION` | `\"video_480p\"` | Base resolution preset (`image_768res` or `video_480p`). |\n\n### Gradio\n```bash\npython lance_gradio_t2v_v2t.py --gpus 0 --server-port 7860\n```\n\n### Benchmarks\n\n#### DPG-Bench Evaluation\n\n\u003Cdiv align=\"center\">\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"left\">Models\u003C\u002Fth>\n      \u003Cth align=\"center\"># Params.\u003C\u002Fth>\n      \u003Cth align=\"center\">Global\u003C\u002Fth>\n      \u003Cth align=\"center\">Entity\u003C\u002Fth>\n      \u003Cth align=\"center\">Attribute\u003C\u002Fth>\n      \u003Cth align=\"center\">Relation\u003C\u002Fth>\n      \u003Cth align=\"center\">Other\u003C\u002Fth>\n      \u003Cth align=\"center\">Overall\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"8\">\u003Ci>Generation-only Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">SDXL\u003C\u002Ftd>\u003Ctd align=\"center\">3.5B\u003C\u002Ftd>\u003Ctd align=\"center\">83.27\u003C\u002Ftd>\u003Ctd align=\"center\">82.43\u003C\u002Ftd>\u003Ctd align=\"center\">80.91\u003C\u002Ftd>\u003Ctd align=\"center\">86.76\u003C\u002Ftd>\u003Ctd align=\"center\">80.41\u003C\u002Ftd>\u003Ctd align=\"center\">74.65\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">DALL-E 3\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">90.97\u003C\u002Ftd>\u003Ctd align=\"center\">89.61\u003C\u002Ftd>\u003Ctd align=\"center\">88.39\u003C\u002Ftd>\u003Ctd align=\"center\">90.58\u003C\u002Ftd>\u003Ctd align=\"center\">89.83\u003C\u002Ftd>\u003Ctd align=\"center\">83.50\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">SD3-Medium\u003C\u002Ftd>\u003Ctd align=\"center\">2B\u003C\u002Ftd>\u003Ctd align=\"center\">87.90\u003C\u002Ftd>\u003Ctd align=\"center\">91.01\u003C\u002Ftd>\u003Ctd align=\"center\">88.83\u003C\u002Ftd>\u003Ctd align=\"center\">80.70\u003C\u002Ftd>\u003Ctd align=\"center\">88.68\u003C\u002Ftd>\u003Ctd align=\"center\">84.08\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">FLUX.1-dev\u003C\u002Ftd>\u003Ctd align=\"center\">12B\u003C\u002Ftd>\u003Ctd align=\"center\">74.35\u003C\u002Ftd>\u003Ctd align=\"center\">90.00\u003C\u002Ftd>\u003Ctd align=\"center\">88.96\u003C\u002Ftd>\u003Ctd align=\"center\">90.87\u003C\u002Ftd>\u003Ctd align=\"center\">88.33\u003C\u002Ftd>\u003Ctd align=\"center\">83.84\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Qwen-Image\u003C\u002Ftd>\u003Ctd align=\"center\">20B\u003C\u002Ftd>\u003Ctd align=\"center\">91.32\u003C\u002Ftd>\u003Ctd align=\"center\">91.56\u003C\u002Ftd>\u003Ctd align=\"center\">92.02\u003C\u002Ftd>\u003Ctd align=\"center\">94.31\u003C\u002Ftd>\u003Ctd align=\"center\">92.73\u003C\u002Ftd>\u003Ctd align=\"center\">88.32\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"8\">\u003Ci>Unified Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Janus-Pro-7B\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">86.90\u003C\u002Ftd>\u003Ctd align=\"center\">88.90\u003C\u002Ftd>\u003Ctd align=\"center\">89.40\u003C\u002Ftd>\u003Ctd align=\"center\">89.32\u003C\u002Ftd>\u003Ctd align=\"center\">89.48\u003C\u002Ftd>\u003Ctd align=\"center\">84.19\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">OmniGen2\u003C\u002Ftd>\u003Ctd align=\"center\">4B\u003C\u002Ftd>\u003Ctd align=\"center\">88.81\u003C\u002Ftd>\u003Ctd align=\"center\">88.83\u003C\u002Ftd>\u003Ctd align=\"center\">90.18\u003C\u002Ftd>\u003Ctd align=\"center\">89.37\u003C\u002Ftd>\u003Ctd align=\"center\">90.27\u003C\u002Ftd>\u003Ctd align=\"center\">83.57\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Show-o2\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">89.00\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>91.78\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">89.96\u003C\u002Ftd>\u003Ctd align=\"center\">91.81\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>91.64\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">86.14\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">BAGEL\u003Csup>†\u003C\u002Fsup>\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">88.94\u003C\u002Ftd>\u003Ctd align=\"center\">90.37\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>91.29\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">90.82\u003C\u002Ftd>\u003Ctd align=\"center\">88.67\u003C\u002Ftd>\u003Ctd align=\"center\">85.07\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">InternVL-U\u003C\u002Ftd>\u003Ctd align=\"center\">1.7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>90.39\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">90.78\u003C\u002Ftd>\u003Ctd align=\"center\">90.68\u003C\u002Ftd>\u003Ctd align=\"center\">90.29\u003C\u002Ftd>\u003Ctd align=\"center\">88.77\u003C\u002Ftd>\u003Ctd align=\"center\">85.18\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">TUNA\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>90.42\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>91.68\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">90.94\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>91.87\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>90.73\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>86.76\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">TUNA-2\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">89.50\u003C\u002Ftd>\u003Ctd align=\"center\">91.40\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>92.07\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">91.91\u003C\u002Ftd>\u003Ctd align=\"center\">88.81\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>86.54\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">🌟 \u003Cb>Lance (Ours)\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>3B\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>83.89\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>91.07\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>89.36\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>93.38\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>80.80\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>84.67\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\u003Cem>\u003Csup>†\u003C\u002Fsup> indicates methods that use LLM rewriters for prompt rewriting before generation.\u003C\u002Fem>\u003C\u002Fp>\n\n#### GenEval Evaluation\n\n\u003Cdiv align=\"center\">\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"left\">Models\u003C\u002Fth>\n      \u003Cth align=\"center\"># Params.\u003C\u002Fth>\n      \u003Cth align=\"center\">1-Obj.\u003C\u002Fth>\n      \u003Cth align=\"center\">2-Obj.\u003C\u002Fth>\n      \u003Cth align=\"center\">Count\u003C\u002Fth>\n      \u003Cth align=\"center\">Colors\u003C\u002Fth>\n      \u003Cth align=\"center\">Position\u003C\u002Fth>\n      \u003Cth align=\"center\">Attr.\u003C\u002Fth>\n      \u003Cth align=\"center\">Overall\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"9\">\u003Ci>Generation-only Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">SDXL\u003C\u002Ftd>\u003Ctd align=\"center\">3.5B\u003C\u002Ftd>\u003Ctd align=\"center\">0.98\u003C\u002Ftd>\u003Ctd align=\"center\">0.74\u003C\u002Ftd>\u003Ctd align=\"center\">0.39\u003C\u002Ftd>\u003Ctd align=\"center\">0.85\u003C\u002Ftd>\u003Ctd align=\"center\">0.15\u003C\u002Ftd>\u003Ctd align=\"center\">0.23\u003C\u002Ftd>\u003Ctd align=\"center\">0.55\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">DALL-E 3\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">0.96\u003C\u002Ftd>\u003Ctd align=\"center\">0.87\u003C\u002Ftd>\u003Ctd align=\"center\">0.47\u003C\u002Ftd>\u003Ctd align=\"center\">0.83\u003C\u002Ftd>\u003Ctd align=\"center\">0.43\u003C\u002Ftd>\u003Ctd align=\"center\">0.45\u003C\u002Ftd>\u003Ctd align=\"center\">0.67\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">SD3-Medium\u003C\u002Ftd>\u003Ctd align=\"center\">2B\u003C\u002Ftd>\u003Ctd align=\"center\">0.99\u003C\u002Ftd>\u003Ctd align=\"center\">0.94\u003C\u002Ftd>\u003Ctd align=\"center\">0.72\u003C\u002Ftd>\u003Ctd align=\"center\">0.89\u003C\u002Ftd>\u003Ctd align=\"center\">0.33\u003C\u002Ftd>\u003Ctd align=\"center\">0.60\u003C\u002Ftd>\u003Ctd align=\"center\">0.74\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">FLUX.1-dev\u003C\u002Ftd>\u003Ctd align=\"center\">12B\u003C\u002Ftd>\u003Ctd align=\"center\">0.98\u003C\u002Ftd>\u003Ctd align=\"center\">0.93\u003C\u002Ftd>\u003Ctd align=\"center\">0.75\u003C\u002Ftd>\u003Ctd align=\"center\">0.93\u003C\u002Ftd>\u003Ctd align=\"center\">0.68\u003C\u002Ftd>\u003Ctd align=\"center\">0.65\u003C\u002Ftd>\u003Ctd align=\"center\">0.82\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Qwen-Image\u003C\u002Ftd>\u003Ctd align=\"center\">20B\u003C\u002Ftd>\u003Ctd align=\"center\">0.99\u003C\u002Ftd>\u003Ctd align=\"center\">0.92\u003C\u002Ftd>\u003Ctd align=\"center\">0.89\u003C\u002Ftd>\u003Ctd align=\"center\">0.88\u003C\u002Ftd>\u003Ctd align=\"center\">0.76\u003C\u002Ftd>\u003Ctd align=\"center\">0.77\u003C\u002Ftd>\u003Ctd align=\"center\">0.87\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"9\">\u003Ci>Unified Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Janus-Pro-7B\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.99\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">0.89\u003C\u002Ftd>\u003Ctd align=\"center\">0.59\u003C\u002Ftd>\u003Ctd align=\"center\">0.90\u003C\u002Ftd>\u003Ctd align=\"center\">0.79\u003C\u002Ftd>\u003Ctd align=\"center\">0.66\u003C\u002Ftd>\u003Ctd align=\"center\">0.80\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">OmniGen2\u003C\u002Ftd>\u003Ctd align=\"center\">4B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>1.00\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">0.95\u003C\u002Ftd>\u003Ctd align=\"center\">0.64\u003C\u002Ftd>\u003Ctd align=\"center\">0.88\u003C\u002Ftd>\u003Ctd align=\"center\">0.55\u003C\u002Ftd>\u003Ctd align=\"center\">0.76\u003C\u002Ftd>\u003Ctd align=\"center\">0.80\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Show-o2\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>1.00\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">0.87\u003C\u002Ftd>\u003Ctd align=\"center\">0.58\u003C\u002Ftd>\u003Ctd align=\"center\">0.92\u003C\u002Ftd>\u003Ctd align=\"center\">0.52\u003C\u002Ftd>\u003Ctd align=\"center\">0.62\u003C\u002Ftd>\u003Ctd align=\"center\">0.76\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">BAGEL\u003Csup>†\u003C\u002Fsup>\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">0.98\u003C\u002Ftd>\u003Ctd align=\"center\">0.95\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.84\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.95\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">0.78\u003C\u002Ftd>\u003Ctd align=\"center\">0.77\u003C\u002Ftd>\u003Ctd align=\"center\">0.88\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Mogao\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>1.00\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.97\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.83\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">0.93\u003C\u002Ftd>\u003Ctd align=\"center\">0.84\u003C\u002Ftd>\u003Ctd align=\"center\">0.80\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.89\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">InternVL-U\u003C\u002Ftd>\u003Ctd align=\"center\">1.7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.99\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">0.94\u003C\u002Ftd>\u003Ctd align=\"center\">0.74\u003C\u002Ftd>\u003Ctd align=\"center\">0.91\u003C\u002Ftd>\u003Ctd align=\"center\">0.77\u003C\u002Ftd>\u003Ctd align=\"center\">0.74\u003C\u002Ftd>\u003Ctd align=\"center\">0.85\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">TUNA\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>1.00\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.97\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">0.81\u003C\u002Ftd>\u003Ctd align=\"center\">0.91\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.88\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.83\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.90\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">TUNA-2\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.99\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>0.96\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">0.80\u003C\u002Ftd>\u003Ctd align=\"center\">0.91\u003C\u002Ftd>\u003Ctd align=\"center\">0.84\u003C\u002Ftd>\u003Ctd align=\"center\">0.76\u003C\u002Ftd>\u003Ctd align=\"center\">0.87\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">🌟 \u003Cb>Lance (Ours)\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>3B\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>1.00\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.94\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.84\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.97\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.87\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.81\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>0.90\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\u003Cem>\u003Csup>†\u003C\u002Fsup> indicates methods that use LLM rewriters for prompt rewriting before generation.\u003C\u002Fem>\u003C\u002Fp>\n\n#### GEdit-Bench Evaluation\n\n\u003Cdiv align=\"center\">\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"left\">Models\u003C\u002Fth>\n      \u003Cth align=\"center\"># Params.\u003C\u002Fth>\n      \u003Cth align=\"center\">BC\u003C\u002Fth>\n      \u003Cth align=\"center\">CA\u003C\u002Fth>\n      \u003Cth align=\"center\">MM\u003C\u002Fth>\n      \u003Cth align=\"center\">MC\u003C\u002Fth>\n      \u003Cth align=\"center\">PB\u003C\u002Fth>\n      \u003Cth align=\"center\">ST\u003C\u002Fth>\n      \u003Cth align=\"center\">SA\u003C\u002Fth>\n      \u003Cth align=\"center\">SR\u003C\u002Fth>\n      \u003Cth align=\"center\">SRp\u003C\u002Fth>\n      \u003Cth align=\"center\">TM\u003C\u002Fth>\n      \u003Cth align=\"center\">TT\u003C\u002Fth>\n      \u003Cth align=\"center\">Avg\u002FG_O\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"14\">\u003Ci>Generation-only Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Gemini 2.0\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">6.32\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">GPT Image 1\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">6.96\u003C\u002Ftd>\u003Ctd align=\"center\">6.85\u003C\u002Ftd>\u003Ctd align=\"center\">7.10\u003C\u002Ftd>\u003Ctd align=\"center\">5.41\u003C\u002Ftd>\u003Ctd align=\"center\">6.74\u003C\u002Ftd>\u003Ctd align=\"center\">7.44\u003C\u002Ftd>\u003Ctd align=\"center\">7.51\u003C\u002Ftd>\u003Ctd align=\"center\">8.73\u003C\u002Ftd>\u003Ctd align=\"center\">8.55\u003C\u002Ftd>\u003Ctd align=\"center\">8.45\u003C\u002Ftd>\u003Ctd align=\"center\">8.69\u003C\u002Ftd>\u003Ctd align=\"center\">7.49\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Qwen-Image-Edit\u003C\u002Ftd>\u003Ctd align=\"center\">20B\u003C\u002Ftd>\u003Ctd align=\"center\">8.23\u003C\u002Ftd>\u003Ctd align=\"center\">8.30\u003C\u002Ftd>\u003Ctd align=\"center\">7.33\u003C\u002Ftd>\u003Ctd align=\"center\">8.05\u003C\u002Ftd>\u003Ctd align=\"center\">7.49\u003C\u002Ftd>\u003Ctd align=\"center\">6.74\u003C\u002Ftd>\u003Ctd align=\"center\">8.57\u003C\u002Ftd>\u003Ctd align=\"center\">8.09\u003C\u002Ftd>\u003Ctd align=\"center\">8.29\u003C\u002Ftd>\u003Ctd align=\"center\">8.48\u003C\u002Ftd>\u003Ctd align=\"center\">8.50\u003C\u002Ftd>\u003Ctd align=\"center\">8.01\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"center\" colspan=\"14\">\u003Ci>Unified Models\u003C\u002Fi>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Lumina-DiMOO\u003C\u002Ftd>\u003Ctd align=\"center\">8B\u003C\u002Ftd>\u003Ctd align=\"center\">3.43\u003C\u002Ftd>\u003Ctd align=\"center\">4.27\u003C\u002Ftd>\u003Ctd align=\"center\">3.08\u003C\u002Ftd>\u003Ctd align=\"center\">2.77\u003C\u002Ftd>\u003Ctd align=\"center\">4.74\u003C\u002Ftd>\u003Ctd align=\"center\">5.19\u003C\u002Ftd>\u003Ctd align=\"center\">4.44\u003C\u002Ftd>\u003Ctd align=\"center\">3.80\u003C\u002Ftd>\u003Ctd align=\"center\">4.38\u003C\u002Ftd>\u003Ctd align=\"center\">2.68\u003C\u002Ftd>\u003Ctd align=\"center\">4.20\u003C\u002Ftd>\u003Ctd align=\"center\">3.91\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Ovis-U1\u003C\u002Ftd>\u003Ctd align=\"center\">1.2B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.49\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">6.88\u003C\u002Ftd>\u003Ctd align=\"center\">6.21\u003C\u002Ftd>\u003Ctd align=\"center\">4.79\u003C\u002Ftd>\u003Ctd align=\"center\">5.98\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>6.46\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">7.49\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.25\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.27\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">4.48\u003C\u002Ftd>\u003Ctd align=\"center\">6.31\u003C\u002Ftd>\u003Ctd align=\"center\">6.42\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">BAGEL\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">7.32\u003C\u002Ftd>\u003Ctd align=\"center\">6.91\u003C\u002Ftd>\u003Ctd align=\"center\">6.38\u003C\u002Ftd>\u003Ctd align=\"center\">4.75\u003C\u002Ftd>\u003Ctd align=\"center\">4.57\u003C\u002Ftd>\u003Ctd align=\"center\">6.15\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.90\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">7.16\u003C\u002Ftd>\u003Ctd align=\"center\">7.02\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.32\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">6.22\u003C\u002Ftd>\u003Ctd align=\"center\">6.52\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">InternVL-U\u003C\u002Ftd>\u003Ctd align=\"center\">1.7B\u003C\u002Ftd>\u003Ctd align=\"center\">7.08\u003C\u002Ftd>\u003Ctd align=\"center\">7.05\u003C\u002Ftd>\u003Ctd align=\"center\">6.38\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.02\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>6.03\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">6.27\u003C\u002Ftd>\u003Ctd align=\"center\">7.13\u003C\u002Ftd>\u003Ctd align=\"center\">6.55\u003C\u002Ftd>\u003Ctd align=\"center\">6.33\u003C\u002Ftd>\u003Ctd align=\"center\">6.59\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>6.85\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">6.66\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">InternVL-U (w\u002F CoT)\u003C\u002Ftd>\u003Ctd align=\"center\">1.7B\u003C\u002Ftd>\u003Ctd align=\"center\">7.05\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.87\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>6.50\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">6.99\u003C\u002Ftd>\u003Ctd align=\"center\">5.77\u003C\u002Ftd>\u003Ctd align=\"center\">6.10\u003C\u002Ftd>\u003Ctd align=\"center\">7.33\u003C\u002Ftd>\u003Ctd align=\"center\">7.16\u003C\u002Ftd>\u003Ctd align=\"center\">7.12\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.36\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">6.46\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>6.88\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">🌟 \u003Cb>Lance (Ours)\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>3B\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.73\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.74\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.28\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.83\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.50\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.03\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>7.64\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.85\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.71\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">4.46\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.57\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>7.30\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n#### VBench Evaluation (Video Generation)\n\n\u003Cdiv align=\"center\">\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"left\">Type\u003C\u002Fth>\n      \u003Cth align=\"left\">Model\u003C\u002Fth>\n      \u003Cth align=\"center\"># Params.\u003C\u002Fth>\n      \u003Cth align=\"center\">Total Score ↑\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd align=\"center\" rowspan=\"12\">\u003Ci>Gen. Only\u003C\u002Fi>\u003C\u002Ftd>\n      \u003Ctd align=\"left\">ModelScope\u003C\u002Ftd>\u003Ctd align=\"center\">1.7B\u003C\u002Ftd>\u003Ctd align=\"center\">75.75\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">LaVie\u003C\u002Ftd>\u003Ctd align=\"center\">3B\u003C\u002Ftd>\u003Ctd align=\"center\">77.08\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Show-1\u003C\u002Ftd>\u003Ctd align=\"center\">6B\u003C\u002Ftd>\u003Ctd align=\"center\">78.93\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">AnimateDiff-V2\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">80.27\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">VideoCrafter-2.0\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">80.44\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">CogVideoX\u003C\u002Ftd>\u003Ctd align=\"center\">5B\u003C\u002Ftd>\u003Ctd align=\"center\">81.61\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Kling\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">81.85\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Open-Sora-2.0\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">81.71\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Gen-3\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">82.32\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Step-Video-T2V\u003C\u002Ftd>\u003Ctd align=\"center\">30B\u003C\u002Ftd>\u003Ctd align=\"center\">81.83\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Hunyuan Video\u003C\u002Ftd>\u003Ctd align=\"center\">-\u003C\u002Ftd>\u003Ctd align=\"center\">83.43\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Wan2.1-T2V\u003C\u002Ftd>\u003Ctd align=\"center\">14B\u003C\u002Ftd>\u003Ctd align=\"center\">83.69\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"center\" rowspan=\"6\">\u003Ci>Unified\u003C\u002Fi>\u003C\u002Ftd>\n      \u003Ctd align=\"left\">HaproOmni\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">78.10\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Emu3\u003C\u002Ftd>\u003Ctd align=\"center\">8B\u003C\u002Ftd>\u003Ctd align=\"center\">80.96\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">VILA-U\u003C\u002Ftd>\u003Ctd align=\"center\">7B\u003C\u002Ftd>\u003Ctd align=\"center\">74.01\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">Show-o2\u003C\u002Ftd>\u003Ctd align=\"center\">2B\u003C\u002Ftd>\u003Ctd align=\"center\">81.34\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">TUNA\u003C\u002Ftd>\u003Ctd align=\"center\">1.5B\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cu>84.06\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd align=\"left\">🌟 \u003Cb>Lance (Ours)\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>3B\u003C\u002Fb>\u003C\u002Ftd>\u003Ctd align=\"center\">\u003Cb>85.11\u003C\u002Fb>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n#### Running Benchmarks\n\nReady-to-run benchmark scripts are provided under `benchmarks\u002F`:\n\n| Benchmark              | Modality | Script                                                        |\n|------------------------|----------|---------------------------------------------------------------|\n| GenEVAL (image gen)    | Image    | `benchmarks\u002Fimage_gen\u002FGenEVAL\u002Fsample_GenEVAL.sh`              |\n| DPG (image gen)        | Image    | `benchmarks\u002Fimage_gen\u002FDPG\u002Fsample_DPG.sh`                      |\n| GEdit (image edit)     | Image    | `benchmarks\u002Fimage_gen\u002FGEdit\u002Fsample_GEdit.sh`                  |\n| VBench (video gen)     | Video    | `benchmarks\u002Fvideo_gen\u002FVbench\u002Fsample_vbench.sh`                |\n\n\n## 📄 License\n\nCopyright 2025 Bytedance Ltd. and\u002For its affiliates.\n\n## 🙏 Acknowledgements\n\nWe would like to thank the contributors of [BAGEL](https:\u002F\u002Fgithub.com\u002FByteDance-Seed\u002Fbagel), [Qwen2.5-VL-3B-Instruct](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2.5-VL-3B-Instruct), and [Wan2.2](https:\u002F\u002Fgithub.com\u002FWan-Video\u002FWan2.2) for their open research and contributions.\n\n## 💖 Citation\n\nIf you find **Lance** useful for your project or research, welcome to 🌟 this repo and cite our work using the following BibTeX:\n\n```bibtex\n@misc{fu2026lanceunifiedmultimodalmodeling,\n      title         = {Lance: Unified Multimodal Modeling by Multi-Task Synergy},\n      author        = {Fengyi Fu and Mengqi Huang and Shaojin Wu and Yunsheng Jiang and Yufei Huo and Hao Li and Yinghang Song and Fei Ding and Jianzhu Guo and Qian He and Zheren Fu and Zhendong Mao and Yongdong Zhang},\n      year          = {2026},\n      eprint        = {2605.18678},\n      archivePrefix = {arXiv},\n      primaryClass  = {cs.CV},\n      url           = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.18678},\n}\n```\n\n## 📞 Contact\n\nFor questions, issues, or collaborations, please contact [Mengqi Huang](https:\u002F\u002Fcorleone-huang.github.io\u002F) and [Jianzhu Guo](https:\u002F\u002Fguojianzhu.com\u002F).\n","Lance 是一个轻量级的统一多模态模型，支持图像和视频的理解、生成与编辑。该项目的核心功能包括在30亿参数规模下高效处理图像生成、图像编辑及视频生成等任务，并且完全从零开始训练而成，仅需128个A100 GPU即可完成。Lance采用分阶段多任务配方构建，适用于需要在一个框架内同时处理多种视觉内容理解与创造的应用场景，如创意设计工具、社交媒体平台的内容自动化生产等。其基于Python开发，遵循Apache License 2.0开源协议。",2,"2026-06-11 03:55:10","CREATED_QUERY"]