Text-to-Image Generation on GenEval long prompts LLM-refined
47.8GenEval ScoreGEAR-1.0B
Evaluation Results
| Method | Links | |
|---|---|---|
| GEAR-1.0BCFG=w/2026.06 | 47.8 | |
| LlamaGen-REPA-1.0BCFG=w/2026.06 | 41.9 | |
| GEAR-1.0BCFG=w/o2026.06 | 22.7 | |
| LlamaGen-REPA-1.0BCFG=w/o2026.06 | 21.8 |