1: \begin{abstract}

2: We consider the  nonparametric robust estimation problem for regression models

3:  in continuous time with  semi-Markov noises. An adaptive model selection

4: procedure  is proposed. Under general moment conditions on the noise distribution 

5: a sharp non-asymptotic oracle inequality for the robust risks is obtained

6:  and the robust efficiency  is shown. It turns out that for semi-Markov models  the

7:  robust   minimax convergence  rate may be faster or slower  than the classical one.

8:  \end{abstract}

9: