{"id":"arxiv/2606.01596","url":"https://arxiv.org/abs/2606.01596","tags":["math.NA","cs.LG"],"score":0,"title":"Second-order consistency for learning chaotic dynamics via randomized Jacobian matching","domain":"math.NA","source":"arxiv","authors":["Shinhoo Kang","Hai V. Nguyen","Tan Bui-Thanh"],"reviews":0,"subject":"math.NA","summary":"Short-horizon accuracy does not ensure that a learned chaotic system has correct long-time dynamics. Trajectory (zeroth-order) matching constrains vector-field values, and Jacobian (first-order) matching constrains local tangent dynamics, but neither determines how the Jacobian varies away from supervised states, so a model can be locally accurate while drifting toward spurious attractors and distorting long-time statistics. We show that second-order supervision mitigates these failures. Because forming full Hessian tensors is computationally prohibitive in high dimensions, we propose…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-01","reactions":{}}