Research map / 3D vision / 3D generation
Towards Fine-Grained Text-to-3D Quality Assessment: A Benchmark and A Two-Stage Rank-Learning Metric
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 prompt-number (PN) curriculum (ablation) | Rank2Score · T23D-CompBench AP (stage 1 only) | SRCC: 0.806 -> 0.838 | Improved |
| add score-difference (SD) curriculum (ablation) | Rank2Score · T23D-CompBench AP (stage 1 only) | SRCC: 0.806 -> 0.826 | Improved |
| add dimension-consistency (DC) curriculum (ablation) | Rank2Score · T23D-CompBench AP (stage 1 only) | SRCC: 0.806 -> 0.825 | Improved |
| combine PN, SD and DC curricula (ablation) | Rank2Score · T23D-CompBench AP (stage 1 only) | SRCC: 0.806 -> 0.858 | Improved |
| add supervised contrastive loss L_cons (ablation) | Rank2Score · T23D-CompBench AP | SRCC: 0.833 -> 0.858 | Improved |
| add second-stage MSE loss on MOS (ablation) | Rank2Score · T23D-CompBench AP | SRCC: 0.858 -> 0.908 | Improved |
| learnable level prompt instead of regression head (ablation) | Rank2Score · T23D-CompBench AP | SRCC: 0.875 -> 0.908 | Improved |
| learnable instead of fixed level prompt (ablation) | Rank2Score · T23D-CompBench AP | SRCC: 0.899 -> 0.908 | Improved |
| place learnable level prompt in middle vs begin or end (ablation) | Rank2Score · T23D-CompBench AP | SRCC: 0.904 -> 0.908 | No clear effect |
| Rank2Score vs best fine-tuned baseline HyperScore | Rank2Score · T23D-CompBench AP | SRCC: 0.809 -> 0.908 | Improved |
| Rank2Score vs fine-tuned ImageReward | Rank2Score · T23D-CompBench AP | SRCC: 0.777 -> 0.908 | Improved |
| Rank2Score vs HyperScore on MATE-3D | Rank2Score · MATE-3D OQ | SRCC: 0.792 -> 0.897 | Improved |
Closest papers by meaning
- Gen3DEval: Using vLLMs for Automatic Evaluation of Generated 3D Objects (CVPR 2025 · similarity 0.89)
- MARVEL-40M+: Multi-Level Visual Elaboration for High-Fidelity Text-to-3D Content Creation (CVPR 2025 · similarity 0.89)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.88)
- GPT-4V(ision) is a Human-Aligned Evaluator for Text-to-3D Generation (CVPR 2024 · similarity 0.88)
- Revisiting text-to-image evaluation with Gecko: on metrics, prompts, and human rating (ICLR 2025 · similarity 0.88)
- Retrieval-Augmented Score Distillation for Text-to-3D Generation (ICML 2024 · 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.
