NVS benchmarks are misaligned: proper pose alignment puts 3D Gaussians back at SOTA
kwangmoo_yi · x · 2026-10-31
The authors announce "Aligning Misaligned Benchmarks for Single-View Novel View Synthesis" (Tseng et al., arXiv:2610.38349). Key finding: test-time optimization for aligning a method's reference frame with ground-truth poses is ill-posed due to a complex optimization landscape, often landing in arbitrary local optima and producing misleading benchmark numbers.
Instead they use traditional 2D-to-2D and 2D-to-3D image matching for alignment, and evaluate only within co-visible regions. After proper alignment, explicit 3D Gaussian methods reach state of the art — reversing prior reports that pure image-based 2D methods outperformed them.
More from Research
- Aligning image and text embeddings without paired data: Wasserstein Procrustes and rotation papers — phillip_isola · 2026-10-31
- Phillip Isola demos aligning image and text embeddings with just a single rotation — phillip_isola · 2026-10-31
- Arena Offers Up to $50,000 Per Project for Academic Research on AI Evaluation — arena · 2026-10-31
- StudentBench: AI tutoring matches human GRE gains at 918x lower cost — gharik · 2026-10-31
- WISD: Distill Training Signal from Weaker Quantized Copies of Reasoning Models — jet-ai · 2026-10-31
- Albert Gu: H-Net learns to segment Chinese characters from raw bytes mid-training — niloofar_mire · 2026-10-31