# 任务 × 主贡献矩阵

口径：[推断] 按 taxonomy v1 的单一主任务和主贡献计数，75篇各计一次；3篇主任务不确定另列，不强行分配。每格列最多3篇，少于3篇则列实际数量。代表论文按机制相关性/易读性人工优选，并非质量排名。次贡献不计入本表，见 papers.csv。

|任务 / 主贡献|表示|条件注入|架构与生成范式|记忆与上下文|训练信号|推理时方法|数据|评测|合计|
|---|---|---|---|---|---|---|---|---|---|
|文本或图像生成视频|0|0|3 · [P064 · CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer](https://arxiv.org/html/2408.06072)；[P066 · Pyramidal Flow Matching for Efficient Video Generative Modeling](https://arxiv.org/html/2410.05954)；[P010 · HunyuanVideo: A Systematic Framework For Large Video Generative Models](https://arxiv.org/html/2412.03603)|0|14 · [P025 · Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion](https://arxiv.org/html/2506.08009)；[P032 · Learning to Credit the Right Steps: Objective-aware Process Optimization for Visual Generation](https://arxiv.org/html/2604.19234)；[P072 · DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models](https://arxiv.org/html/2506.03517)|2 · [P060 · Physical Simulator In-the-Loop Video Generation](https://arxiv.org/html/2603.06408)；[P071 · Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search](https://arxiv.org/html/2501.19252)|2 · [P021 · SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset](https://arxiv.org/html/2406.14477)；[P065 · OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation](https://arxiv.org/html/2407.02371)|2 · [P020 · VBench: Comprehensive Benchmark Suite for Video Generative Models](https://arxiv.org/html/2311.17982)；[P057 · T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation](https://arxiv.org/html/2407.14505)|23|
|条件控制生成|2 · [P008 · MIMO: Controllable Character Video Synthesis with Spatial Decomposed Modeling](https://arxiv.org/html/2409.16160)；[P014 · Any2Caption:Interpreting Any Condition to Caption for Controllable Video Generation](https://arxiv.org/html/2503.24379)|8 · [P005 · MotionCtrl: A Unified and Flexible Motion Controller for Video Generation](https://arxiv.org/html/2312.03641)；[P006 · CameraCtrl: Enabling Camera Control for Video Diffusion Models](https://arxiv.org/html/2404.02101)；[P056 · Mind the Time: Temporally-Controlled Multi-Event Video Generation](https://arxiv.org/html/2412.05263)|1 · [P001 · Video-to-Video Synthesis](https://arxiv.org/html/1808.06601)|0|2 · [P013 · OmniHuman-1: Rethinking the Scaling-Up of One-Stage Conditioned Human Animation Models](https://arxiv.org/html/2502.01061)；[P016 · VideoMage: Multi-Subject and Motion Customization of Text-to-Video Diffusion Models](https://arxiv.org/html/2503.21781)|1 · [P007 · PhysGen: Rigid-Body Physics-Grounded Image-to-Video Generation](https://arxiv.org/html/2409.18964)|0|0|14|
|编辑与修改|1 · [P059 · VideoCoF: Unified Video Editing with Temporal Reasoner](https://arxiv.org/html/2512.07469)|3 · [P067 · Unified In-Context Video Editing](https://arxiv.org/html/2506.04216)；[P052 · Precise Editing and Flexible Referencing for Interactable Worlds](https://arxiv.org/html/2609.34470)；[P061 · ReCamMaster: Camera-Controlled Generative Rendering from A Single Video](https://arxiv.org/html/2503.11647)|1 · [P019 · Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation](https://arxiv.org/html/2212.11565)|1 · [P074 · PermaVid: Consistent Video Generation Across Edits via Disentangled Context Memory](https://arxiv.org/html/2606.16449)|1 · [P002 · Dreamix: Video Diffusion Models are General Video Editors](https://arxiv.org/html/2302.01329)|0|3 · [P018 · TrajectoryMover: Generative Movement of Object Trajectories in Videos](https://arxiv.org/html/2603.29092)；[P058 · FFP-300K: Scaling First-Frame Propagation for Generalizable Video Editing](https://arxiv.org/html/2601.01720)；[P070 · Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists](https://arxiv.org/pdf/2502.06734)|0|10|
|长视频与多镜头叙事|0|0|1 · [P053 · WorldAttention: An Efficient Attention Architecture for Interactive Video World Models](https://arxiv.org/pdf/2609.34606)|1 · [P069 · LongLive: Real-time Interactive Long Video Generation](https://arxiv.org/html/2509.22622)|0|0|0|1 · [P075 · EntityBench: Towards Entity-Consistent Long-Range Multi-Shot Video Generation](https://arxiv.org/html/2605.15199)|3|
|动作条件的预测与交互世界模型|2 · [P040 · Genie: Generative Interactive Environments](https://arxiv.org/html/2402.15391)；[P043 · AdaWorld: Learning Adaptable World Models with Latent Actions](https://arxiv.org/html/2503.18938)|2 · [P017 · Hand2World: Autoregressive Egocentric Interaction Generation via Free-Space Hand Gestures](https://arxiv.org/html/2602.09600)；[P045 · Ctrl-World: A Controllable Generative World Model for Robot Manipulation](https://arxiv.org/html/2510.10125)|5 · [P041 · Diffusion Models Are Real-Time Game Engines](https://arxiv.org/html/2408.14837)；[P068 · Horizon Imagination: Efficient On-Policy Rollout in Diffusion World Models](https://arxiv.org/html/2602.08032)；[P037 · Learning to Simulate Dynamic Environments with GameGAN](https://arxiv.org/html/2005.12126)|2 · [P044 · WorldMem: Long-term Consistent World Simulation with Memory](https://arxiv.org/html/2504.12369)；[P050 · Echo-Memory: A Controlled Study of Memory in Action World Models](https://arxiv.org/html/2606.09803)|5 · [P049 · ActionSplice: In-Flight Action Editing for Interactive World Models](https://arxiv.org/html/2609.08230)；[P048 · Training Agents Inside of Scalable World Models](https://arxiv.org/html/2509.24527)；[P038 · Learning Interactive Real-World Simulators](https://arxiv.org/html/2310.06114)|0|0|0|16|
|3D/4D 一致生成|0|0|0|2 · [P062 · VMem: Consistent Interactive Video Scene Generation with Surfel-Indexed View Memory](https://arxiv.org/html/2506.18903)；[P015 · GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control](https://arxiv.org/html/2503.03751)|0|0|0|1 · [P063 · WorldScore: A Unified Evaluation Benchmark for World Generation](https://arxiv.org/html/2504.00983)|3|
|视频作为推理或回答|0|0|0|0|2 · [P034 · Wan-R1: Verifiable-Reinforcement Learning for Video Reasoning](https://arxiv.org/html/2603.27866)；[P036 · Video Models Can Reason with Verifiable Rewards](https://arxiv.org/html/2605.15458)|0|0|1 · [P030 · Respond Beyond Language: A Benchmark for Video Generation in Response to Realistic User Intents](https://arxiv.org/html/2506.01689)|3|
|其他（说明）|0|0|0|0|0|0|0|0|0|
|不确定|0|1 · [P011 · VACE: All-in-One Video Creation and Editing](https://arxiv.org/html/2503.07598)|0|1 · [P051 · WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory](https://arxiv.org/html/2607.02517)|0|0|0|1 · [P073 · WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors](https://arxiv.org/html/2605.10434)|3|

总数75；其中72篇有确定主任务。“其他”0篇；不确定3篇（VACE、WorldDirector、WorldReasonBench）。完整主次贡献见CSV。