Comment on “Multispectral radiation temperature inversion based on Transformer-LSTM-SVM”

A recent publication (Cui et al., 2025) presents an inversion method to extract temperature from multiwavelength radiation measurements that is claimed to be accurate despite the absence of any information on emissivity. This Comment recalls that when no information on emissivity is available, we are faced with an unsolvable problem because it has a continuous infinity of physically acceptable solutions, among which the true solution is indistinguishable from the others. It is therefore impossible to extract reliably the true temperature. No inversion or optimization method can be successful, except by chance. However, in Cui et al. (2025), the validation tests showed systematic success, which can then only be explained by the fact that the “true” solution was known in advance. When this is not the case, the inversion method cannot be better than a random draw. Blind tests of multiwavelength thermometry methods allegedly not requiring emissivity information were recently organized. The results confirmed, if proof were needed, that this approach is a dead end.