iCreatAI News
Skip to content
Trending storyWatching

Apple proposes inference-time scaling of latent 3D representations

1 reports1 sources2 days agoUpdated

Understand this story

AI overview

The Apple research team proposes the split-and-scale method: at inference time, a single pretrained tokenizer is applied to arbitrarily small spatial subregions of a scene, thereby scaling the capacity of latent 3D representations without increasing training cost. The team says extensive ablation experiments validate the trade-offs among compute, memory, latent capacity, and reconstruction quality, with reconstruction quality competitive with larger models that exceed single-GPU memory limits. The team also trained an autoregressive 3D generator based on this method and says its performance is comparable to current state-of-the-art baselines.

AI overview based on source reports · Updated 2 hr ago

Coverage timeline

Follow the coverage to explore different perspectives.

Nov 16
  1. Apple Machine Learning Research
    Split-and-Scale Latent 3D Representations

    The authors propose split-and-scale, an inference-time method that expands latent capacity by applying a single pretrained tokenizer to arbitrarily small sub-regions of a 3D scene, since representing scenes as continuous functions of 3D coordinates lets the same latent capacity cover a smaller spatial extent and capture far higher detail. They conduct ablations on trade-offs between compute, memory, latent capacity, and reconstruction quality, showing the method achieves quality competitive with much larger models that exceed single-GPU memory limits. They also train an autoregressive 3D generator with split-and-scale that performs competitively with state-of-the-art baselines.

Activity for this story

Not enough continuous observations to show a trend yet.