Research map / Image generation / Text-to-image diffusion
LongT2IBench: A Benchmark for Evaluating Long Text-to-Image Generation with Graph-structured Annotations
Zhichao Yang, Tianjiao Gu, Jianjie Wang, Feiyu Lin, Xiangfei Sheng, Pengfei Chen, Leida Li
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 Hierarchical Alignment CoT (ablation) | LongT2IExpert · LongT2IBench scoring | Avg. (SRCC+PLCC)/2: 0.516 -> 0.557 | Improved |
| add Hierarchical Alignment CoT (ablation) | LongT2IExpert · LongT2IBench interpreting | Overall accuracy: 39.9% -> 53.2% | Improved |
| add interpretation task in multi-task training (ablation) | LongT2IExpert · LongT2IBench scoring | Avg. (SRCC+PLCC)/2: 0.474 -> 0.557 | Improved |
| add scoring task in multi-task training (ablation) | LongT2IExpert · LongT2IBench interpreting | Overall accuracy: 32.8% -> 53.2% | Improved |
| fine-tune baselines on LongT2IBench training data | PickScore · LongT2IBench scoring (overall) | Avg. (SRCC+PLCC)/2: 0.336 -> 0.438 | Improved |
| fine-tune baselines on LongT2IBench training data | ImageReward · LongT2IBench scoring (overall) | Avg. (SRCC+PLCC)/2: 0.227 -> 0.439 | Improved |
| fine-tune baselines on LongT2IBench training data | FGA-BLIP2 · LongT2IBench scoring (overall) | Avg. (SRCC+PLCC)/2: 0.246 -> 0.425 | Improved |
| increase prompt length to 110+ words | VLM-based evaluators (CLIP-style encoders) · LongT2IBench scoring, 110+ words | SRCC: ImageReward 0.283 -> -0.023 (30-50 to 110+ words) | Hurt |
| use MLLM-based evaluator instead of VLM-based at 110+ words | Q-Eval-Score · LongT2IBench scoring, 110+ words | SRCC: Q-Eval-Score 0.422 vs HPSv2 0.148 | Improved |
| longer prompts lower LongT2IExpert correlation | LongT2IExpert · LongT2IBench scoring | SRCC: 0.781 (30-50 words) -> 0.431 (110+ words) | Hurt |
Closest papers by meaning
- Revisiting text-to-image evaluation with Gecko: on metrics, prompts, and human rating (ICLR 2025 · similarity 0.90)
- Discriminative Probing and Tuning for Text-to-Image Generation (CVPR 2024 · similarity 0.90)
- DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models (NeurIPS 2026 · similarity 0.90)
- Improving Long-Text Alignment for Text-to-Image Diffusion Models (ICLR 2025 · similarity 0.89)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.89)
- Aligning Text to Image in Diffusion Models is Easier Than You Think (NeurIPS 2025 · similarity 0.88)
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.
