Text-to-Score Generation on Custom Text-to-Score prompt set 1.0 (test) benchmark leaderboard · SOTA2 Research

Text-to-Score Generation on Custom Text-to-Score prompt set 1.0 (test)

100Valid File Generation RateMidi-LLM

Evaluation Results

Method	Links
Midi-LLM2026.05		100	—	—	95.77	69.96	95.31	92.81	90.31	79.83	62.73	84.82	89.08	79.04	100	39.66	48.1	50.46	0.0825	84.24	81.01	2.4
Infer-Align2026.05		99.58	—	—	96.34	89.43	97.33	98.35	95.65	89.46	95.15	82.36	92.47	90.1	45.38	25.21	47.9	23.82	0.0207	38.79	36.69	2.3
Text2Score2026.05		99.16	2	238	97.58	98.16	99.67	99.6	98.57	96.48	97.07	92.5	99.19	96.04	92.61	84.78	100	83.76	0.1446	73.81	70.96	3.07
MidiLM2026.05		97.9	—	—	93.73	86.74	97.44	95.48	93.93	82.56	87.36	84.88	80.8	84.47	90.56	56.65	97.42	41.39	0.0935	49.19	44.37	2.09
ComposerX2026.05		50	91.56	4,484	85.11	91	98.74	90.06	90.29	98.51	83.3	98.4	99.54	95.22	99.16	100	100	55.07	0.1266	90.27	88.55	2.34