Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Split-and-Scale Expands 3D Representation Capacity at Inference

AI summary

Apple Machine Learning Research lists a NeurIPS paper introducing split-and-scale, a method that expands latent capacity at inference while keeping training costs manageable. The method represents 3D scenes as continuous functions of 3D coordinates, allowing one pretrained tokenizer to operate on small scene sub-regions with more detail. The authors report competitive reconstruction and generative-modeling results, including an autoregressive 3D generator competitive with state-of-the-art baselines.

In-depth insight

The approach addresses the memory and compute costs of increasing latent capacity during training by scaling detailed 3D representations at inference, as reported by the paper’s authors.

research area Computer Visionconference NeurIPS

content type paperpublished November 2026

Split-and-Scale Latent 3D Representations

AuthorsDorian Chan, Muhammed Kocabas, Xiaoming Zhao, Oncel Tuzel, Jen-Hao Rick Chang

View publication

Latent 3D representations have significantly expanded the capabilities of modern shape modeling, enabling compact encoding of 3D scenes and supporting powerful 3D generative models. Despite this progress, current representations struggle to capture the fine details of complex objects, fundamentally limited by the capacity of their latents. However, scaling latent capacity at training time is computationally expensive and often infeasible under memory limits, and naively increasing the number of latents at inference fails to make effective use of the added capacity. In this paper, we address both challenges with split-and-scale, a method that effectively expands latent capacity at inference while keeping training cost tractable. Our key observation is that by representing 3D scenes as continuous functions of 3D coordinates, a single pretrained tokenizer can be applied at inference to arbitrarily small sub-regions of a scene, allocating the same latent capacity to a smaller spatial extent and thus capturing far higher detail. We validate this insight by conducting extensive ablations to study the trade-offs between compute, memory, latent capacity, and reconstruction quality, demonstrating our inference-time scaling method achieves quality competitive with much larger models that exceed single-GPU memory limits. We further demonstrate that split-and-scale supports high-quality generative modeling, training an autoregressive 3D generator that performs competitively with state-of-the-art baselines.

#Apple#NeurIPS

Original · 2 min readRead the original
Group chat

No comments yet. Start the conversation.