AI Paper Atlas

Research map / Benchmarks / PIE-Bench

PIE-Bench leaderboard

23 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on PIE-Bench. Best PSNR: 29.44 (Ours (SD3), 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.

PIE-Bench · PSNR

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

#PSNR ↑MethodPaperTopic
129.44Ours (SD3)
SD3
Multi-History-Step SDE Inversion for Image Editing with Superior Regional Awareness
ECCV 2026 · Table 1
Editing and style transfer
229.2LOCATEdit
PnP-I, DDIM(50)
LOCATEdit: Graph Laplacian Optimized Cross Attention for Localized Text-Guided Image Editing
ICCV 2025 · Table 1
Text-to-image diffusion
328.33Navi-FlowAlign (M=1)Semantic Granularity Navigation in Image Editing
ICML 2026 · Table 1
Editing and style transfer
427.95EditInfinity
Infinity-2B
EditInfinity: Image Editing with Binary-Quantized Generative Models
NeurIPS 2025 · Table 1
Editing and style transfer
527.85PnP Inversion + AutoEdit
SD 1.5
AutoEdit: Automatic Hyperparameter Tuning for Image Editing
NeurIPS 2025 · Main results table
Editing and style transfer
627.46Ours
SD3.5
Wavelet-Guided Semantic Signal Compensation for Inversion-Free Image Editing
ECCV 2026 · Appendix table
Editing and style transfer
727.32FIA-Edit
SD3.5
FIA-Edit: Frequency-Interactive Attention for Efficient and High-Fidelity Inversion-Free Text-Guided Image Editing
AAAI 2026 · Table 1
Editing and style transfer
826.97DRFS
SD3.5
Delta Rectified Flow Sampling for Text-to-Image Editing
CVPR 2026 · Table 1
Editing and style transfer
926.67BitResEdit
Infinity-2B
Edit the Bits, Diff the Codes: Bitwise Residual Editing for Visual Autoregressive Models
NeurIPS 2026 · Table 1
Editing and style transfer
1026.66DNAEdit
SD3.5
DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing
NeurIPS 2025 · Table: Quantitative comparison
Editing and style transfer
11 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

PIE-Bench · CLIP similarity (edited)

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

#CLIP similarity (edited) ↑MethodPaperTopic
123.83DRFS
SD3
Delta Rectified Flow Sampling for Text-to-Image Editing
CVPR 2026 · Table 1
Editing and style transfer
223.83LTP + VFA
SD3
SplitFlow: Flow Decomposition for Inversion-Free Text-to-Image Editing
NeurIPS 2025 · Table 2
Editing and style transfer
323.02LOCATEdit
PnP-I, DDIM(50)
LOCATEdit: Graph Laplacian Optimized Cross Attention for Localized Text-Guided Image Editing
ICCV 2025 · Table 1
Text-to-image diffusion
422.87DNAEdit
FLUX
DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing
NeurIPS 2025 · Table: Quantitative comparison
Editing and style transfer
522.82FIA-Edit
SD3.5
FIA-Edit: Frequency-Interactive Attention for Efficient and High-Fidelity Inversion-Free Text-Guided Image Editing
AAAI 2026 · Table 1
Editing and style transfer
622.77AREdit (Ours)
VAR
Training-Free Text-Guided Image Editing with Visual Autoregressive Model
ICCV 2025 · Table 1
Editing and style transfer
722.67DirectEdit
SD3.5-medium
DirectEdit: Step-Level Accurate Inversion for Flow-Based Image Editing
ICML 2026 · Table 1
Editing and style transfer
822.56Ours (theta=0.2, RF-inv)
Flux.1-dev
Exploring Multimodal Diffusion Transformers for Enhanced Prompt-based Image Editing
ICCV 2025 · Table A1
Text-to-image diffusion
922.31Janus-Pro-R1-Edit
Janus-Pro-R1
Janus-Pro-R1: Advancing Collaborative Visual Comprehension and Generation via Reinforcement Learning
NeurIPS 2025 · Table: image editing results
Multimodal LLMs
1021.88SCAR
LlamaGen-XL
Semantic Context Matters: Improving Conditioning for Autoregressive Models
CVPR 2026 · Table 3
Autoregressive and latent 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

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