Flame3D: Zero-shot Compositional Reasoning of 3D Scenes with Agentic Language Models

TL;DR

Flame3D is a training-free framework enabling zero-shot, compositional reasoning about 3D scenes by using external tools and large language models, without requiring 3D-specific training.

Contribution

It introduces Flame3D, a novel inference-time approach that constructs editable 3D scene memories and synthesizes spatial programs for open-ended reasoning.

Findings

Competitive performance on ScanQA without training

Essential role of synthesized spatial operations in reasoning

Effective reasoning over layouts and objects not in the scene

Abstract

3D scene understanding spans reasoning about free space, object grounding, hypothetical object insertions, complex geometric relationships, and integrating all of these with external tools and data sources. Existing 3D understanding methods typically rely on large-scale 3D-language training or focus on object grounding and simple spatial relationships. We argue that the broad generalization that motivates 3D-language training can be achieved at inference time, without 3D-specific training. We propose Flame3D, a training-free framework that represents scenes as editable visual-textual 3D memories and exposes them to an off-the-shelf MLLM through composable spatial tools. Flame3D also lets the agent synthesize custom spatial programs at inference time, enabling open-ended reasoning over layouts, empty space, and objects not yet present in the scene. External data and corrections can be added to the memory without retraining. In addition to showing competitive performance to finetuned 3D-LMM methods on ScanQA, we study multi-hop 3D reasoning capabilities of Flame3D by evaluating it on a curated compositional spatial-reasoning benchmark, Compose3D. We find that fixed tools fall short and that the agent's ability to synthesize spatial operations at inference time is essential. These results invite the question: should future progress in 3D scene understanding focus on richer scene memories and expressive compositional abstractions?