Enhancing Lightweight Vision Language Models through Group Competitive Learning for Socially Compliant Navigation

TL;DR

This paper introduces Group Competitive Learning (GCL), a novel training strategy that significantly enhances the reasoning and decision-making capabilities of lightweight vision language models for socially compliant navigation, achieving high accuracy with lower computational costs.

Contribution

The paper proposes GCL, a new training approach combining global semantics and distributional regularization, to improve lightweight VLMs' performance in social navigation tasks.

Findings

GCL improves VLM F1 scores by up to 40%.

Lightweight models outperform larger models after GCL training.

GCL enables efficient and accurate social navigation in real-world scenarios.

Abstract

Social robot navigation requires a sophisticated integration of scene semantics and human social norms. Scaling up Vision Language Models (VLMs) generally improves reasoning and decision-making capabilities for socially compliant navigation. However, increased model size incurs substantial computational overhead, limiting suitability for real-time robotic deployment. Conversely, lightweight VLMs enable efficient inference but often exhibit weaker reasoning and decision-making performance in socially complex environments. Achieving both strong reasoning ability and efficiency remains an open challenge. To bridge this gap, we propose Group Competitive Learning (GCL), a strategy designed to amplify the capabilities of lightweight VLMs. Our strategy introduces the Group Competitive Objective (GCO) to harmonize global semantics with distributional regularization, alongside Asymmetric Group Optimization (AGO) to explore the upper limits of model performance. Empirical evaluations on social navigation benchmarks demonstrate that GCL significantly elevates VLM performance. Specifically, GCL enables the Qwen2.5-VL-3B learner model and guide Qwen3-VL-4B to achieve an F1 score of 0.968 and 0.914, representing 40\% and 12\% improvement over vanilla supervised fine-tuning (SFT). Notably, under vanilla SFT, the 3B model initially trails the 8B model (F1: 0.692 vs. 0.755). However, through the GCL, the 3B model outperforms (28\%) the 8B baseline model. These results suggest that GCL provides an effective solution for achieving both high accuracy and computational efficiency in real-world deployment.