A Appendix

A.1 Extended results We benchmark models in terms of retrieval accuracy as described ( 4.3) for both the NIST20 and NPLIB1 datasets (Table A1, A2). We recreate retrieval experiments using the full PubChem retrieval library in Table A3. We reproduce main text results with standard error values included in Tables A4, A5, and A6. We showcase additional spectra predictions from our model trained on NIST20 in Figure A1. All values represent the mean across three separate random seeds the standard error of the mean for a single test set. All values represent the mean across three separate random seeds the standard error of the mean for a single test set.