Search papers, labs, and topics across Lattice.
This paper introduces SpatialOPSD, a self-distillation framework that enables Multimodal Large Language Models (MLLMs) to internalize spatial reasoning capabilities from verified coding agent traces without external tools. By treating agent traces as privileged information and employing Repetition-Aware Distillation to prevent leakage, SpatialOPSD trains a standalone MLLM to perform complex spatial tasks. Experimental results show that SpatialOPSD significantly outperforms supervised fine-tuning and GRPO on both in-distribution and out-of-distribution spatial reasoning benchmarks.
Spatial coding agents can be distilled into standalone MLLMs, achieving superior spatial reasoning and generalization without external tools.
Spatial coding agents significantly improve spatial reasoning in Multimodal Large Language Models (MLLMs) by using external tools to generate verified execution traces. However, this paradigm inherently suffers from prohibitive inference-time overhead and external dependencies. In this paper, we explore whether an MLLM can internalize this agentic capability to operate entirely tool-free. We begin with a simple observation: prompting an MLLM with summarized execution traces of a spatial coding agent naturally unlocks the model's internal spatial Chain-of-Thought (CoT). Motivated by this, we introduce SpatialOPSD, an on-policy self-distillation framework that internalizes spatial reasoning into a standalone MLLM by formulating verified agent traces as privileged information. To mitigate privileged-information leakage during distillation, we introduce Repetition-Aware Distillation, which combines repetition masking with unlikelihood regularization. Experiments across multiple benchmarks demonstrate that self-distilling SpatialOPSD achieves higher average accuracy than SFT and GRPO on both spatial and OOD datasets, exhibiting superior performance and generalization.
