Text-to-action generation on OpenExp-Action-100 benchmark leaderboard · SOTA2 Research

Text-to-action generation on OpenExp-Action-100

16.33Validity Score (Val)GPT-5

Evaluation Results

Method	Links
GPT-52026.04		16.33	51.55	37.43	52.33	71	51.62	65.47
Gemini-2.5-Pro2026.04		14.67	58.72	44.23	60	71.71	54.24	66.51
GPT-52026.04		13.33	50.61	36.56	—	70.73	51.31	65.06
Gemini-2.5-Pro2026.04		12.67	58.26	43.99	59.67	71.28	54.21	66.49
DeepSeek-R12026.04		10.67	58.99	43.01	61.33	71.62	52.79	65.85
DeepSeek-R12026.04		10.33	58.28	42.37	56.67	71.13	52.31	65.65
DeepSeek-R12026.04		8.67	38.32	26.21	22	59.12	42.32	53.56
GPT-52026.04		1.35	50.39	39.39	50.17	66.31	52.53	62.76
Gemini-2.5-Pro2026.04		0.33	51.78	37.23	37.33	62.92	46.25	58.9