FT-NCFM: An Influence-Aware Data Distillation Framework for Efficient VLA Models

TL;DR

This paper presents FT-NCFM, a data-centric framework that distills valuable training data for VLA models, significantly reducing training time while maintaining high performance.

Contribution

It introduces a novel data distillation framework using a self-contained Fact-Tracing engine to generate a model-agnostic, information-rich data subset for VLA models.

Findings

Models trained on 5% of distilled data achieve 85-90% success rate.

Training time is reduced by over 80%.

The framework outperforms traditional data and model compression methods.

Abstract

The powerful generalization of Vision-Language-Action (VLA) models is bottlenecked by their heavy reliance on massive, redundant, and unevenly valued datasets, hindering their widespread application. Existing model-centric optimization paths, such as model compression (which often leads to performance degradation) or policy distillation (whose products are model-dependent and lack generality), fail to fundamentally address this data-level challenge. To this end, this paper introduces FT-NCFM, a fundamentally different, data-centric generative data distillation framework. Our framework employs a self-contained Fact-Tracing (FT) engine that combines causal attribution with programmatic contrastive verification to assess the intrinsic value of samples. Guided by these assessments, an adversarial NCFM process synthesizes a model-agnostic, information-dense, and reusable data asset. Experimental results on several mainstream VLA benchmarks show that models trained on just 5% of our distilled coreset achieve a success rate of 85-90% compared with training on the full dataset, while reducing training time by over 80%. Our work demonstrates that intelligent data distillation is a highly promising new path for building efficient, high-performance VLA models.