Amin Karimi Monsefi

Amin Karimi Monsefi

Ph.D. candidate at The Ohio State University. Efficient and controllable generative models — diffusion and flow matching across vision and language.

Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation

Amin Karimi Monsefi, Dominic Culver, Nikhil Bhendawade, Manuel R. Ciosici, Yizhe Zhang, Irina Belousova

arXiv, 2026

Figure from Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation

When a distilled few-step student underperforms, the usual explanation is insufficient capacity. This work argues the opposite: the trajectory is the bottleneck, not the student. TS-DFM replaces the blind stochastic jumps of standard trajectory construction with guided navigation, using a lightweight energy compass to evaluate candidate continuations at each midpoint. At 170M parameters the 8-step student reaches 32% lower perplexity than its 1024-step teacher while being 128x faster, with no change to inference cost.

Paper