为论文配图而生的中文 Prompt 库
精选中文学术 prompt：研究框架图、技术路线图、机制示意图、统计图。一键复制、一键试用。
214 prompts
如何写出高质量 Prompt
评分高的 prompt 共享四个要素：先点目标、再列要素、定布局、最后定风格。
1. 目标
一句话说清要画什么。控制在 25 字以内。
2. 要素
用 bullet 列出要包含的模块、阶段、数据系列或分子 —— 模型会按 prompt 原文复现标签。
3. 布局
明确选一个朝向（左到右、上到下、2x2 网格、辐射状），不要混合，否则模型会混乱。
4. 风格
指定调色板、字体、目标期刊（NeurIPS、Nature、IEEE），以及图生图时的保留规则。
常见问题
浏览全部 Prompt 样例
直达每个分类与精选 Prompt。
文生图
- PINN Framework Pipeline for Microscopy Data
- DP-FTRL Privacy Amplification Analysis — Conference Poster
- NC-TTT: Noise-Contrastive Test-Time Training Poster
- Pref-CTRL: Value Function Training Overview
- CROTTC-IF Architecture Overview
- RC-RAG: Paraphrase-Guided RAG for Relation Completion
- gMatch Method Overview
- CinDM: Compositional Inverse Design Poster
- MOMENTA Architecture: Pipeline and Optimization
- Pose-Guided Person Image Synthesis — CVPR 2024 Poster
- ResAD: Class-Generalizable Anomaly Detection Poster
- PRISM Framework: Multimodal Sentiment Overview
- FastGen: Adaptive KV Cache Compression for LLMs
- SurroundSDF: Vision-Centric 3D Scene Poster
- ProDock Docking Workflow Overview
- Implicit Variant Architecture with InternVL3 and LoRA
- Lateral-Axial Factorization Architecture
- EigenVI: Score-Based Variational Inference Poster
- Leverage SHAP: Shapley Values via Leverage Sampling
- DEPT: Decoupling Embeddings from Transformers — ICLR Poster
- HistDiT: Transformer Diffusion Architecture for Histology
- DAS: Diffusion Alignment as Sampling Poster
- MindBridge: Cross-Subject Brain Decoding Poster
- Relative Camera Pose Estimation via SDP Relaxation
- Fourier Latent Dynamics for Motion Learning — Poster
- CausalRivers: Causal Discovery Benchmark Poster
- Neural Fields Scaling Laws Across Tasks — Poster
- Proposed Framework Overview
- Readout Guidance for Diffusion Control — Poster
- PASR Architecture for 3D Shape Pose Retrieval
- Sketch Animation Pipeline from Text — Poster
- Muscle-Driven Piano Playing System Overview
- MASA: Label-Free Any-Object Tracking — Poster
- HydraLoRA Asymmetric Fine-Tuning Method — Poster
- DEADiff Diffusion for Stylized Generation — Poster
- Unsupervised 3D Occupancy Learning — Poster
- SemJudge HSG Visualization for Ink-Wash Art
- Counterfactual Randomization Sampling — Poster
- Molecular Property Prediction Survey Overview
- DFA Class Bias in Cross-Attention Layers
- Universal Pose Denoiser Transformer Architecture
- GFlowState Visualization Workflow Diagram
- HyPeR Framework for Audio Reasoning Overview
- VPSeg Vanishing-Point Video Segmentation — Poster
- PRDL 3D Face Reconstruction Method — Poster
- UniMODE Unified Monocular 3D Detection — Poster
- FlowVid Consistent Video-to-Video Synthesis — Poster
- MCSC Multimodal Context-to-Script Task Overview
- SocraticLM LLM Teaching Framework — Poster
- Overlapping Genes Coding Constraints Schematic
- GOPSA EEG Domain Adaptation Method — Poster
- ObyGaze12 Visual Objectification in Films — Poster
- Urban Dynamic Scene 3D Gaussians — Poster
- Neural Diffusion for Syntax-Tree Synthesis — Poster
- CLARITY Framework Overview with Rule and LLM Components
- SLLM Training with Bias Word Position Tag Prediction
- Reasoning-Augmented Occupation Prediction Framework
- D-Persona Medical Image Segmentation — Poster
- JKO Scheme for Learning Diffusion Processes — Poster
- Selective Learning Framework Pipeline Overview
- VARIANT Web Server for Viral Mutation Analysis
- LOKI Synthetic Data Detection Benchmark — Poster
- MCR-RNNT Dual-Mode Unified Transducer Training
- JudgeLM Fine-Tuning LLMs as Scalable Judges — Poster
- DLM4G Graph-Conditioned Diffusion Language Framework
- Multi-Grained Video Anomaly Detection — CVPR Poster
- Hammer Function-Calling Models with Masking — Poster
- GROOT Self-Supervised Pre-Training in Minecraft — Poster
- EEG Stop-Signal Task Classification Pipeline
- GeoLink 3D Structure Anchors for Cross-View Geo-Localization
- MDP Sample Complexity Bounds — NeurIPS Poster
- OmniMouse Unified Neural Forecasting Architecture
- Long-Term 3D Dynamics Parsing Pipeline — Poster
- FoL++ Visual Place Recognition Training Pipeline
- Watermarking Makes Language Models Radioactive — Poster
- ArXivDoc LaTeX-Native Scientific Document Retrieval
- HumanGaussian Text-to-3D Human Generation — Poster
- E-SRResNet Super-Resolution Network Architecture
- Projector vs Phoneme Speech-Language Interfaces
- Joint 1-D and 2-D Audio Modeling Framework Schematic
- Dual-Glob Contrastive F0 Contour Framework
- Wonder3D Single-Image to 3D Diffusion — Poster
- SERE Semi-Supervised Dual-Path Architecture for CLSER
- LLM Pipeline for Deriving AV Driving Requirements
- MeshRet Skinned Motion Retargeting — Poster
- CFAT Triangular-Attention Super-Resolution — Poster
- LucidDreamer Text-to-3D with ISM — Poster
- Human Expertise in Algorithmic Prediction — Poster
- Uncertainty Feedback Loop for Camera Pose Precision
- SNAN Spiking Neuron-Astrocyte Grid Navigation Model
- SUE Spatial Uncertainty for Visual Place Recognition
- SeaEvo Evolutionary Loop Framework Overview
- ROSE Scoring Cascade Workflow
- IAM-CL2R Compatible Learning Framework — Poster
- MILLET Interpretable Time Series Classifier — Poster
- Overall Model Architecture Overview
- Multi-Agent Adversarial Synthesis and Defense Pipeline
- AudioHijack Attack Framework
- Pangea Unified Semantic Space for Actions — Poster
- ProciGen Template-Free HOI Reconstruction Poster
- MVFA Visual Feature Adaptation for Medical Anomaly Detection
- Bayesian Model of Haptic Stimulus Perception
- Snap Video Text-to-Video Generation Poster
- LORS Method for Efficient Network Stacking Poster
- ArbGraph Credibility Arbitration Framework Overview
- Total Selfie Full-Body Selfie Generation Poster
- NTK Surrogate Models for Explaining Neural Networks
- RaTA-Tool Multimodal Tool Retrieval Pipeline
- Social Reward Image Scoring Model Poster
- DSPO Score-Matching Alignment for Diffusion Models
- CS-ARM-BN Meta-Learning Method and Comparison
- Tuning Frequency Bias in State Space Models Poster
- Mathematical Framework for Platelet Plug Coagulation
- Sema Token-Based Speech Coding Architecture Comparison
- Hi-WM Human-in-the-World-Model Pipeline
- Document Agentic RAG Workflow Overview
- Behavior Understanding Alignment (BUA) Framework
- Overview of Different Types of Generative Models
- SurFITR MLLM Tampering Generation Pipeline
- Mapping Deep Networks to Human Visual Brain Poster
- TeMuDance Text and Music Driven Dance Generation
- Neologism Learning Study Design Overview for NNS Writers
- Any2Graph: Supervised Graph Prediction Poster
- NASNet Search Space: Repeated Cell Stack Architecture
- Self-Supervised Learning Task Construction Summary
- Compressive Transformer Memory Slots for Long Context
- Vanilla Transformer vs Transformer-XL Training Comparison
- NAS Cell Blocks and Per-Step Operation Decisions
- GPipe Pipeline Parallelism with Microbatches
- LLM.int8() Mixed-Precision Decompositions
- How DeceptionNet Works: Domain Randomization Overview
- Unsupervised Data Generation (UDG) Framework Overview
- SimOpt Framework Overview for Sim-to-Real Transfer
- SAFE for Long-Form LLM Factuality Evaluation
- How to Build Toolformer: Tool-Use Training Pipeline
- Vanilla Transformer Model Architecture
- Reflexion Framework Illustration for LLM Agents
- Two-Stage Bootstrap Toxicity Classifier on Reddit Data
- VAAL: Variational Adversarial Active Learning
- Perspective API Toxicity Scores Overview
- How Algorithm Distillation (AD) Works
- Meta-Learning System: Supervisory and Subordinate RNNs
- Meta-RL with Outer and Inner Optimization Loops
草图重绘
- 5G-Based Real-Time 2D Human Pose Recognition Prototype
- Clarifier Framework for Multimodal Ambiguous Queries
- LiteGE: Shape Descriptor Creation Process
- SSL Representation: Skill Artifact Structuring Overview
- Generative Classifiers for Image and Text Tasks
- GenieReasoner: Unified Embodied Reasoning and Control
- GeoBench Overview: Verified Geometry Reasoning Tasks
- Sync-LoRA Overview for Video Editing Propagation
- Reference-Guided Clustering for Efficient Training
- Conditional Likelihood for Image Detection
- CNSight LLM Pipeline for Clinical Note Sections
- CAMP-VQA Quality-Aware Captioning for Video Quality
- Categorization of Object Tracking Methods
- DINOv2 Resolution-Prediction Network Overview
- CAGenMol Condition-Aware Molecule Diffusion Overview
- Zero-Shot Learning-Based Data Splitting Strategy
- UltraShape 1.0 VAE Pipeline Overview
- FitControler Fit-Aware Virtual Try-On Pipeline
- TIPPo Framework Overview
- BatteryAgent Three-Layer Diagnosis Framework
- Language-Embedded 3D Gaussians for Large-Scale Scenes
- DART and DRIFT Diffusion Priors for MRI and 4D-STEM
- MultiDx Overall Architecture
- PoseFM Flow-Based Camera Pose Estimation Framework
- Object-Aware 2D Gaussian Splatting Pipeline Overview
- IK-GAT Kinematic Graph Attention Architecture
- LVLM vs LLM Instruction-Following Ability Gap
- Three-Level vs Fixed Encryption Strategy Comparison
- VistaBot Architecture for 4D Geometry and Policy Execution
- AV-Dialog Audio-Visual Turn-Taking Overview
- SkinGPT-4 System Design and Interactive Diagnosis
- DiffHDR Framework for LDR-to-HDR Video Reconstruction
- MINT-Bench Taxonomy and Evaluation Pipeline Overview
- AICoFe System Classroom Peer Assessment Setup
- Holistic Cognitive Development (HCD) Framework
关键元素
- GR-Dexter: Dexterous Long-Horizon Task Learning
- NIRNL Framework: CMP and NIR Module Overview
- NeuroSym-BO: Bayesian Optimization Closed Loop
- INR-Based STEM Tomographic Reconstruction Overview
- Spatial Intelligence Pre-Training Survey Structure
- Proposed Method Framework Illustration
- Robotic Tactile Sensing System for Fabric Quality
- PianoFlow Model Architecture with Control Inputs
- Sensory Source Filter and Transform Pipeline
- ARCAS AR Headset and LiDAR Positioning System
- LSRE Latent World Model for Semantic Risk Estimation
- FrameDiffuser Architecture with Dual Conditioning
- CREPES-X Robocentric Relative State Estimation System
- SimPhony Photonic Cross-Layer Co-Exploration Framework
- Mimic2DM View-Agnostic 2D Motion Tracking Pipeline
- Promptable Multi-Granularity Image-Text Embeddings
- vMDM Motion Representation Diffusion Framework
- Methodology for Training Language Model Forecasters
- Multimodal Fusion Framework for Remote Sensing
- Three-Module Perception, Decision, and Authority System
- JEPA-WM World Model Training and Planning Pipeline
- Multi-View Material Reconstruction Pipeline With 3DGS
- HaineiFRDM Frame Restoration Training Pipeline
- LumiX Text-to-Intrinsic Map Generation Overview
- LouvreSAE Training and Inference Stages
- Embodied Manipulation Survey Overview
- SoulX-FlashTalk Performance Overview and Applications
局部重绘
- Geometric Style Transfer Method Overview
- Repository Generation Evaluation Design and Metrics
- HalluAudio Framework for LALM Hallucination Analysis
- TASU2 Framework Overview
- VulGD Dynamic Data Pipeline Workflow
- FacTool Framework for Multi-Task Factuality Evaluation
- Architectures for Modeling Multiple Annotator Labels
- Frozen Model Training and Testing Architecture
- Sequential CNN and Recurrent Cell Representations