Distilling Routed 3D Privilege for Spatial Reasoning in Vision-Language Models

arXiv:2610.12355 · cs.CV, cs.AI · Submitted 2026-10-08 · Read on arXiv

cs.CV, cs.AI

Submitted: 2026-10-08

Updated: 2026-10-08

Code: https://github.com/ZJU-REAL/GPD

Terminology

• GRPO — 44× in this paper · explained in 6 episodes
• Distillation — 27× in this paper · explained in 2 episodes
• Self-distillation — 13× in this paper · explained in 6 episodes
• Benchmark — 9× in this paper · explained in 2 episodes
• Accuracy — 8× in this paper · explained in 1 episode
• Ablation — 7× in this paper · explained in 1 episode
• Computer Vision and Pattern Recognition — 5× in this paper
• Fine-Tuning — 2× in this paper · explained in 5 episodes
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• World Model — 2× in this paper · explained in 7 episodes
• Ablation Studies — 1× in this paper · explained in 1 episode
• Ablation study — 1× in this paper · explained in 1 episode
• Adaptation — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Core — 1× in this paper · explained in 2 episodes
• GRPO Objective — 1× in this paper · explained in 2 episodes
• Grounding — 1× in this paper · explained in 3 episodes
• Knowledge Distillation — 1× in this paper · explained in 11 episodes
• Multimodal Reasoning — 1× in this paper · explained in 5 episodes
• On-policy Self-Distillation (OPSD) — 1× in this paper · explained in 2 episodes
• Regularization — 1× in this paper · explained in 3 episodes
• Trace — 1× in this paper · explained in 2 episodes
• Vision-Language Model (VLM) — 1× in this paper · explained in 3 episodes

Sources

• Qwen3-VL Technical Report
• Qwen2.5-VL Technical Report
• Thinking Without Images: Internalizing Visual Manipulation with On-Policy Self-Distillation
• SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning
• TIGeR: Tool-Integrated Geometric Reasoning in Vision-Language Models for Robotics
• Distilling the Knowledge in a Neural Network
• GPT-4o System Card
• Rethinking On-Policy Self-Distillation for Thinking Models
• Hindsight Compacts but Does Not Repair: Rethinking On-Policy Self-Distillation in Reasoning Models
• LLaVA-OneVision: Easy Visual Task Transfer
• ViewSpatial-Bench: Evaluating Multi-perspective Spatial Localization in Vision-Language Models
• SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models
• Visual-OPSD: Cross-Modal On-Policy Self-Distillation for Efficient Unified Multimodal Reasoning
• Depth Anything 3: Recovering the Visual Space from Any Views
• SpatialReasoner: Towards Explicit and Generalizable 3D Spatial Reasoning
• SpaceR: Reinforcing MLLMs in Video Spatial Reasoning
• GPT4Scene: Understand 3D Scenes from Videos with Vision-Language Models
• Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• Purified OPSD: On-Policy Self-Distillation Without Losing How to Think

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models