发表机构
Phillips Exeter Academy; Cheung Kong Graduate School of Business（菲利普斯埃克塞特学院; 长江商学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文通过对照实验，在LiDAR语义场景补全中发现，迭代优化仅能在连贯缺口场景下提升性能，广泛稀疏化需靠训练覆盖率，且无法修复加性杂波，明确了测试时优化的适用边界。
AI 中文摘要
补全模型应在测试时通过迭代消耗额外计算资源，还是将相似的参数预算用于更宽的单次预测器？去噪课程、损坏增强、容量及非配对评估极易混淆该问题的答案。本文在LiDAR语义场景补全中研究此问题，对比单次预测器、参数匹配的更宽预测器，以及从同一冻结预测器初始化的权重绑定多网格优化器。该方案分离了连贯区域移除、独立稀疏化、距离相关衰减与加性杂波，同时保留精确的场景条件配对。在5个训练种子和815个SemanticKITTI序列-08帧上，完整迭代系统在连续角度移除下，较宽对照的mIoU提升0.911个点，95%移动块自举区间为[0.804, 1.040]，超过预先声明的0.5个点实用阈值；在独立75%稀疏化下，迭代仅提升0.300个点[0.166, 0.436]，而观测族增强提升5.975个点[5.662, 6.140]；两种干预均无法修复加性杂波。迭代系统每帧耗时10.74ms、显存占用0.75GiB，而宽对照为6.25ms、0.23GiB。这些结果确立了几何条件下的经验边界而非通用优势：连贯缺口可证明固定深度优化的合理性，广泛稀疏化的证据更适合通过训练覆盖率解决，而虚假证据需不同的鲁棒性机制。
英文摘要
Should a completion model spend extra test-time compute by iterating, or spend a similar parameter budget on a wider one-shot predictor? The answer is easily confounded by denoising curricula, corruption augmentation, capacity, and unpaired evaluation. We study this question in LiDAR semantic scene completion by comparing a one-shot predictor, a parameter-matched wider predictor, and a weight-tied multigrid refiner initialized from the same frozen predictor. The protocol separates coherent region removal, independent thinning, range-dependent attenuation, and additive clutter while preserving exact scene-condition pairing. Across five training seeds and 815 SemanticKITTI sequence-08 frames, the full iterative system improves mIoU over the wide control by 0.911 points under contiguous angular removal, with a 95% moving-block bootstrap interval of [0.804, 1.040] that clears a predeclared 0.5-point practical margin. Under independent 75% thinning, iteration adds only 0.300 points [0.166, 0.436], whereas observation-family augmentation adds 5.975 points [5.662, 6.140]. Neither intervention repairs additive clutter. The iterative system also costs 10.74 ms and 0.75 GiB per frame, versus 6.25 ms and 0.23 GiB for the wide control. These results establish a geometry-conditioned empirical boundary rather than a universal advantage: coherent gaps can justify fixed-depth refinement, broadly thinned evidence is addressed more effectively by training coverage, and spurious evidence requires a different robustness mechanism.
Comments24 pages, 11 figures, 3 tables
