Research map / Image generation / Text-to-image diffusion
T2I-Copilot: A Training-Free Multi-Agent Text-to-Image System for Enhanced Prompt Interpretation and Interactive Generation
Chieh-Yun Chen, Min Shi, Gong Zhang, Humphrey Shi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add T2I-Copilot agents on top of FLUX.1-dev | FLUX.1-dev · GenAI-Bench advanced | VQAScore: 0.646 -> 0.747 | Improved |
| add T2I-Copilot agents on top of FLUX.1-dev | FLUX.1-dev · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.745 -> 0.813 | Improved |
| add Input Interpreter agent (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.755 -> 0.813 | Improved |
| add Input Interpreter agent (ablation) | T2I-Copilot · GenAI-Bench advanced | VQAScore: 0.646 -> 0.747 | Improved |
| add Quality Evaluator agent (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.805 -> 0.813 | Improved |
| add Quality Evaluator agent (ablation) | T2I-Copilot · GenAI-Bench advanced | VQAScore: 0.736 -> 0.747 | Improved |
| add human feedback to the Quality Evaluator | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.813 -> 0.839 | Improved |
| replace GPT-4o-mini with Mistral Small 3.1-24B (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.813 -> 0.821 | Improved |
| replace GPT-4o-mini with Qwen2.5-VL-7B (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.813 -> 0.811 | No clear effect |
| replace GPT-4o-mini with Qwen2.5-VL-3B (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: 0.813 -> 0.777 | Hurt |
| add RAG-Diffusion as extra generation model (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: -3.43% | Hurt |
| add IP-Adapter as extra reference-based model (ablation) | T2I-Copilot · GenAI-Bench (1600 prompts, overall) | VQAScore: -0.39% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GenArtist — GenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editing (NeurIPS 2024) · better on 3 of 3 shared results
- Lumina-Image-2.0 — Lumina-Image 2.0: A Unified and Efficient Image Generative Framework (ICCV 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Preference Adaptive and Sequential Text-to-Image Generation (ICML 2025 · similarity 0.90)
- PromptCoT: Align Prompt Distribution via Adapted Chain-of-Thought (CVPR 2024 · similarity 0.90)
- DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models (NeurIPS 2026 · similarity 0.90)
- Ranni: Taming Text-to-Image Diffusion for Accurate Instruction Following (CVPR 2024 · similarity 0.90)
- MARS: Mixture of Auto-Regressive Models for Fine-grained Text-to-image Synthesis (AAAI 2025 · similarity 0.90)
- Rich Human Feedback for Text-to-Image Generation (CVPR 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
