AI Paper Atlas

Research map / Benchmarks / GenEval

GenEval leaderboard

161 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on GenEval. Best overall: 0.98 (SymbOmni, ECCV 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.

GenEval · overall

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

#overall ↑MethodPaperTopic
10.98SymbOmniSymbOmni: Evolving Agentic Omni Models via Symbolic Concept Learning
ECCV 2026 · Table 2
Text-to-image diffusion
20.96LatentMorph
Janus-Pro
Show, Don't Tell: Morphing Latent Reasoning into Image Generation
ICML 2026 · Table 1
Text-to-image diffusion
30.96SANA-1.5 4.8B Ours + Inference Scaling
SANA-1.5 4.8B Ours
SANA 1.5: Efficient Scaling of Training-Time and Inference-Time Compute in Linear Diffusion Transformer
ICML 2025 · Detailed GenEval table
Fast diffusion sampling
40.95OmniGen2
Qwen2.5-VL-3B + DiT
OmniGen2: Towards Instruction-Aligned Multimodal Generation
CVPR 2026 · Table 2
Multimodal LLMs
50.94VisionDirector
Qwen-Image
VisionDirector: Vision-Language Guided Closed-Loop Refinement for Generative Image Synthesis
CVPR 2026 · Table 4
Multimodal LLMs
60.94RAISERAISE: Requirement-Adaptive Evolutionary Refinement for Training-Free Text-to-Image Alignment
CVPR 2026 · Table 1
Text-to-image diffusion
70.92LatentUM-Vis-Gen + pixel-rewardLatentUM: Unleashing the Potential of Interleaved Cross-Modal Reasoning via a Latent-Space Unified Model
NeurIPS 2026 · Table 2
Multimodal LLMs
80.91SPAR-3BSPAR: Semantic-Pixel Self-Alignment and Adaptive Routing for Unified Multimodal Models
ECCV 2026 · Table 3
Multimodal LLMs
90.91ReflectionFlow (FLUX.1-dev, 32 samples)From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning
ICCV 2025 · Table 1
Text-to-image diffusion
100.9UniLIP-3BUniLiP: Adapting CLIP for Unified Multimodal Understanding, Generation and Editing
ICLR 2026 · Table 4
Multimodal LLMs
79 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

GenEval · overall score

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

#overall score ↑MethodPaperTopic
10.97PromptRL w/ PE
FLUX.1-dev
PromptRL: Prompt Matters in RL for Flow-Based Image Generation
ICML 2026 · Table 3
Text-to-image diffusion
20.97DiffusionCompass (1K iter)
SD3.5-M
On-Policy Diffusion Reinforcement Learning Meets Off-Policy Quality Anchoring
ECCV 2026 · Table 1a
RL from feedback
30.96BAGEL + Ours (R3)
BAGEL
Understanding vs. Generation: Navigating Optimization Dilemma in Multimodal Models
ICLR 2026 · GenEval table
Multimodal LLMs
40.96URSA (w/ UDM-GRPO)
URSA 1.7B
UDM-GRPO: Stable and Efficient Group Relative Policy Optimization for Uniform Discrete Diffusion Models
ICML 2026 · Table 1
RL from feedback
50.96DMSampler
SD3.5-M
Distillation Models are Good Samplers for Diffusion Reinforcement Learning
ICML 2026 · Table 3
RL from feedback
60.96OctoT2IOctoT2I: A Self-Evolving Agentic Text-to-Image Router
CVPR 2026 · Table 1
Text-to-image diffusion
70.96CGPO
SD3.5-M
Curriculum Group Policy Optimization: Adaptive Sampling for Unleashing the Potential of Text-to-Image Generation
CVPR 2026 · Table 1
Text-to-image diffusion
80.95UniDFlow (4B)Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow Matching
ICML 2026 · Table 2
Multimodal LLMs
90.92Lumina-DiMOO + HT-GRPO
Lumina-DiMOO
Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models
NeurIPS 2026 · Table 1
RL from feedback
100.9091+ RubricRL
Qwen-Image 4B (diffusion)
RubricRL: Simple Generalizable Rewards for Text-to-Image Generation
ECCV 2026 · Table 1/3
Text-to-image diffusion
37 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

GenEval · Counting

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

#Counting ↑MethodPaperTopic
10.95OctoT2IOctoT2I: A Self-Evolving Agentic Text-to-Image Router
CVPR 2026 · Table 1
Text-to-image diffusion
20.95RAISERAISE: Requirement-Adaptive Evolutionary Refinement for Training-Free Text-to-Image Alignment
CVPR 2026 · Table 1
Text-to-image diffusion
30.9GCPO
Janus-Pro-7B
Group Critical-token Policy Optimization for Autoregressive Image Generation
ICLR 2026 · Table 1
Autoregressive and latent generation
40.9ReflectionFlow (FLUX.1-dev, 32 samples)From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning
ICCV 2025 · Table 1
Text-to-image diffusion
50.84STARFlow2 (Ours)
Qwen2.5-VL-7B-Instruct
STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation
NeurIPS 2026 · Table 2
Human and embodied modeling
60.81MMaDA w/ RebRL
MMaDA-8B-Base
RebRL: Reinforcing Discrete Visual Diffusion Models with Rebalanced Timestep Credits
CVPR 2026 · Table 1
RL from feedback
70.8RAGD
Flux.1-dev
RAGD: Regional-Aware Diffusion Model for Text-to-Image Generation
ICCV 2025 · Table 1
Text-to-image diffusion
80.78PromptEnhancer (PE)
Qwen-Image
PromptEnhancer: Taming Your Rewriter for Text-to-Image Generation via Fine-Grained Reward
CVPR 2026 · Table 3
Text-to-image diffusion
90.78ParaUniParaUni: Enhance Generation in Unified Multimodal Model with Reinforcement-driven Hierarchical Parallel Information Interaction
CVPR 2026 · Table 1
Multimodal LLMs
100.77Ours (JS)
Flux-Dev
Designing Instance-Level Sampling Schedules via REINFORCE with James-Stein Shrinkage
CVPR 2026 · Table 3
Text-to-image diffusion
8 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

GenEval · Position

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

#Position ↑MethodPaperTopic
11OctoT2IOctoT2I: A Self-Evolving Agentic Text-to-Image Router
CVPR 2026 · Table 1
Text-to-image diffusion
20.96SANA-1.5 4.8B Ours + Inference Scaling
SANA-1.5 4.8B Ours
SANA 1.5: Efficient Scaling of Training-Time and Inference-Time Compute in Linear Diffusion Transformer
ICML 2025 · Detailed GenEval table
Fast diffusion sampling
30.93ReflectionFlow (FLUX.1-dev, 32 samples)From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning
ICCV 2025 · Table 1
Text-to-image diffusion
40.92UniGen-1.5
Qwen2.5-7B
UniGen-1.5: Enhancing Image Generation and Editing through Reward Unification in RL
CVPR 2026 · Table 2
Text-to-image diffusion
50.86STARFlow2 (Ours)
Qwen2.5-VL-7B-Instruct
STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation
NeurIPS 2026 · Table 2
Human and embodied modeling
60.85Query-KontextQuery-Kontext: An Unified Multimodal Model for Image Generation and Editing
ECCV 2026 · Table 3
Multimodal LLMs
70.83ParaUniParaUni: Enhance Generation in Unified Multimodal Model with Reinforcement-driven Hierarchical Parallel Information Interaction
CVPR 2026 · Table 1
Multimodal LLMs
80.83RAISERAISE: Requirement-Adaptive Evolutionary Refinement for Training-Free Text-to-Image Alignment
CVPR 2026 · Table 1
Text-to-image diffusion
90.83MMaDA w/ RebRL
MMaDA-8B-Base
RebRL: Reinforcing Discrete Visual Diffusion Models with Rebalanced Timestep Credits
CVPR 2026 · Table 1
RL from feedback
100.82RAGD
Flux.1-dev
RAGD: Regional-Aware Diffusion Model for Text-to-Image Generation
ICCV 2025 · Table 1
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

GenEval · GenEval score

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

#GenEval score ↑MethodPaperTopic
10.9725TP-GRPO (w constraint)
SD3.5-M
Alleviating Sparse Rewards by Modeling Step-Wise and Long-Term Sampling Effects in Flow-Based GRPO
ICML 2026 · Table 1
Fast diffusion sampling
20.81MODUS
BAGEL-7B
MODUS: Decoder-only Any-to-Any Modeling of Diverse Modalities
ICML 2026 · Table 1
Cross-modal learning
30.77SD 3.5 + Diff-Aid
SD3.5-Large
Diff-Aid: Inference-time Adaptive Interaction Denoising for Rectified Text-to-Image Generation
NeurIPS 2026 · Table 2
Text-to-image diffusion
40.7UniVGUniVG: A Generalist Diffusion Model for Unified Image Generation and Editing
ICCV 2025 · Table 1
Text-to-image diffusion
50.69Infinity + RADAR
Infinity
RADAR: VQ-VAE Decoder of VAR is a Good Student for Restoring Against Degradation by Acceleration
CVPR 2026 · Table 2
Autoregressive and latent generation
60.69Ours (HPP+PWP+Q)
SD3.5 Large Turbo
HierarchicalPrune: Position-Aware Compression for Large-Scale Diffusion Models
AAAI 2026 · Table 1
Fast diffusion sampling
70.6755FK (k=4)
SDXL-DPO
A General Framework for Inference-time Scaling and Steering of Diffusion Models
ICML 2025 · Table 2
Fast diffusion sampling
80.65OneDiffusionOne Diffusion to Generate Them All
CVPR 2025 · Table 1
Fast diffusion sampling
90.63Transfusion 7B, 2T tokensTransfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model
ICLR 2025 · Table: comparison with image g
Human and embodied modeling
100.496PRISM
SD-1.5
Long-Text-to-Image Generation via Compositional Prompt Decomposition
ICLR 2026 · Appendix table
Text-to-image diffusion
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

GenEval · score

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

#score ↑MethodPaperTopic
10.89Wan-WeaverWan-Weaver: Interleaved Multi-modal Generation via Decoupled Training
CVPR 2026 · Table 2
Multimodal LLMs
20.88Ours (w/ Coord)
Bagel
ConsistCompose: Unified Multimodal Layout Control for Image Composition
CVPR 2026 · Table 3a
Text-to-image diffusion
30.81Bifrost-1 (Qwen2.5-VL 7B)
Qwen2.5-VL 7B
Bifrost-1: Bridging Multimodal LLMs and Diffusion Models with Patch-level CLIP Latents
NeurIPS 2025 · Table 3
Multimodal LLMs
40.8ARR-RPO (Gemini 3.1 Pro)
FLUX.1-dev
Auto-Rubric as Reward: From Implicit Preference to Explicit Generative Criteria
NeurIPS 2026 · Table 2
Alignment and preferences
50.7Goku-T2I (2B)Goku: Flow Based Video Generative Foundation Models
CVPR 2025 · Table 4
Video generation
60.66SnapGenSnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training
CVPR 2025 · Table 3
Text-to-image diffusion
70.63JanusFlow
DeepSeek-LLM 1.3B
JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation
CVPR 2025 · Table 2
Human and embodied modeling
80.62OmniFlowOmniFlow: Any-to-Any Generation with Multi-Modal Rectified Flows
CVPR 2025 · Table 2
Text-to-image diffusion
90.6Ours
Gemma 2 2B
Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for Text-to-Image Synthesis
CVPR 2025 · Table 6
Text-to-image diffusion
100.58Orthus
Chameleon-7B
Orthus: Autoregressive Interleaved Image-Text Generation with Modality-Specific Heads
ICML 2025 · Table: Comparison with state-o
Human and embodied modeling

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