AI Paper Atlas

Research map / Benchmarks / T2I-CompBench

T2I-CompBench leaderboard

16 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on T2I-CompBench. Best Color: 0.8753 (PerfGuard, ICLR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

T2I-CompBench · Color

16 papers, each with its best reported number for its own method; higher is better.

#Color ↑MethodPaperTopic
10.8753PerfGuardPerfGuard: A Performance-Aware Agent for Visual Content Generation
ICLR 2026 · Table 1
LLM agents
20.8726MUSES
SD 3 ControlNet
Muses: 3D-Controllable Image Generation via Multi-Modal Agent Collaboration
AAAI 2025 · Table: Evaluation Results on T
Text-to-image diffusion
30.8643LayerCraft
FLUX.1-dev
LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration
NeurIPS 2025 · Table: Comparison on T2I-Compb
Text-to-image diffusion
40.8482GenArtistGenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editing
NeurIPS 2024 · Table: Quantitative Comparison
Text-to-image diffusion
50.8442PromptEnhancer (PE)
Qwen-Image
PromptEnhancer: Taming Your Rewriter for Text-to-Image Generation via Fine-Grained Reward
CVPR 2026 · Table 4
Text-to-image diffusion
60.8379SD3.5-M+Flow-GRPO
SD3.5-M
Flow-GRPO: Training Flow Matching Models via Online RL
NeurIPS 2025 · T2I-CompBench++ Result table
RL from feedback
70.8211Lumina-Image 2.0Lumina-Image 2.0: A Unified and Efficient Image Generative Framework
ICCV 2025 · Table 3
Text-to-image diffusion
80.8039RAGD
Flux.1-dev
RAGD: Regional-Aware Diffusion Model for Text-to-Image Generation
ICCV 2025 · Table 1
Text-to-image diffusion
90.7919MAGNET (COCO)
SD3
Seg4Diff: Unveiling Open-Vocabulary Semantic Segmentation in Text-to-Image Diffusion Transformers
NeurIPS 2025 · Table 2
Text-to-image diffusion
100.783SD-Turbo + ReNO
SD-Turbo
ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization
NeurIPS 2024 · Table: Quantitative Results on
Text-to-image diffusion
6 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

T2I-CompBench · Shape

11 papers, each with its best reported number for its own method; higher is better.

#Shape ↑MethodPaperTopic
10.7366PerfGuardPerfGuard: A Performance-Aware Agent for Visual Content Generation
ICLR 2026 · Table 1
LLM agents
20.7046LayerCraft
FLUX.1-dev
LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration
NeurIPS 2025 · Table: Comparison on T2I-Compb
Text-to-image diffusion
30.6948GenArtistGenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editing
NeurIPS 2024 · Table: Quantitative Comparison
Text-to-image diffusion
40.6812MUSES
SD 3 ControlNet
Muses: 3D-Controllable Image Generation via Multi-Modal Agent Collaboration
AAAI 2025 · Table: Evaluation Results on T
Text-to-image diffusion
50.613SD3.5-M+Flow-GRPO
SD3.5-M
Flow-GRPO: Training Flow Matching Models via Online RL
NeurIPS 2025 · T2I-CompBench++ Result table
RL from feedback
60.6032RealCompo
SD v1.5 + GLIGEN
RealCompo: Balancing Realism and Compositionality Improves Text-to-Image Diffusion Models
NeurIPS 2024 · T2I-CompBench results table
Text-to-image diffusion
70.6028Lumina-Image 2.0Lumina-Image 2.0: A Unified and Efficient Image Generative Framework
ICCV 2025 · Table 3
Text-to-image diffusion
80.6016RAGD
Flux.1-dev
RAGD: Regional-Aware Diffusion Model for Text-to-Image Generation
ICCV 2025 · Table 1
Text-to-image diffusion
90.4832SDXL-base + MCCD
SDXL-base
MCCD: Multi-Agent Collaboration-based Compositional Diffusion for Complex Text-to-Image Generation
CVPR 2025 · Table 1
Text-to-image diffusion
100.4832Goku-T2I (2B)Goku: Flow Based Video Generative Foundation Models
CVPR 2025 · Table 4
Video generation
1 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

T2I-CompBench · Texture

11 papers, each with its best reported number for its own method; higher is better.

1 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.