Event

Lingchu Intelligence et al.'s FAFM method accepted to NeurIPS 2026

1 reports · 1 sources

The frequency-aware flow matching method FAFM, proposed by Lingchu Intelligence in collaboration with Beihang University, Peking University, and The Chinese University of Hong Kong, has had its paper accepted to the NeurIPS 2026 main conference. The method uses the discrete cosine transform (DCT-II) to represent actions as frequency-domain coefficients corresponding to real time, performs flow matching in coefficient space and constrains the first-order time derivative, requires no additional learnable parameters, and can be used independently or integrated into VLA models such as π₀ and π₀.₅. The team conducted real-robot experiments on the Franka robotic arm platform and evaluated FAFM's performance on four types of fine manipulation tasks in a robotic surgery simulation environment. Lingchu Intelligence states that on two tasks, the success rate of the original π₀.₅ was 80% in both.