Linguistic Controllability on Alpaca-GPT4 (test) benchmark leaderboard · SOTA2 Research

Linguistic Controllability on Alpaca-GPT4 (test)

64Word Count RatioGPT-4

Evaluation Results

Method	Links
GPT-42024.06		64	52	34	51	63	73	46	39	21	0.32	0.38	0.28	0.49	0.51
GPT-3.52024.06		48	53	65	37	67	60	39	44	40	0.29	0.35	0.28	0.4	0.48
GPT-42024.06		40	40	47	31	46	51	35	36	36	0.28	0.3	0.27	0.4	0.32
GPT-3.52024.06		37	50	36	44	55	34	33	54	26	0.33	0.48	0.29	0.39	0.47
IFT2024.06		33	32	33	26	39	41	31	28	38	0.3	0.33	0.32	0.39	0.31
IFT2024.06		31	33	34	32	38	35	33	35	30	0.3	0.31	0.32	0.34	0.31
RLHF (REINFORCE)2024.06		25	24	27	40	27	25	25	27	29	0.29	0.38	0.3	0.83	0.36
MCTune2024.06		18	22	25	25	21	23	22	25	27	0.26	0.29	0.26	0.31	0.2