SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces

Key points

• 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).1
• 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.1
• To mitigate privileged-information leakage during distillation, we introduce Repetition-Aware Distillation, which combines repetition masking with unlikelihood regularization.1
• 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.1

Sources (2)

• [1]SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces
  Hugging Face Daily Papers · Oct 8, 12:00 AM

  Spatial coding agents significantly improve spatial reasoning in Multimodal Large Language Models (MLLMs) by using external tools to generate verified execution traces.

  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).

• [2]SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 06:58 AM · same content

Extractive summary: sentences quoted from the sources.