Readability Style Transfer on CNN/DM (test) benchmark leaderboard · SOTA2 Research

Readability Style Transfer on CNN/DM (test)

27.42FRE DeltaGPT-4o-mini

Evaluation Results

Method	Links
GPT-4o-mini2026.01		27.42	47.42	60.92	0.14
GPT-4o-mini2026.01		19.04	70.96	54.12	0.24
SFT2026.01		16.03	73.97	61.63	0.29
GPT-4o-mini2026.01		15.44	—	59.31	0.27
SFT2026.01		12.27	32.27	55.22	0.29
SFT + PPO (Rsent + Rcons)2026.01		11.23	78.77	58.62	0.38
SFT + PPO (Rsent + Rlex + Rcons)2026.01		11.13	78.87	58.12	0.38
GPT-4o-mini2026.01		10.73	60.73	62.54	0.3
SFT + PPO (Rlex + Rcons)2026.01		10.04	79.96	56.73	0.4
SFT + PPO (Rsent + Rcons)2026.01		8.87	28.87	53.06	0.32
SFT2026.01		8.81	—	60.54	0.34
SFT + PPO (Rsent + Rlex + Rcons)2026.01		6.67	26.67	51.49	0.33
SFT + PPO (Rsent + Rcons)2026.01		6.39	—	59.02	0.37
SFT + PPO (Rsent + Rlex + Rcons)2026.01		5.86	—	58.3	0.37
SFT + PPO (Rlex + Rcons)2026.01		5.71	25.71	49.99	0.34
SFT + PPO (Rlex + Rcons)2026.01		5.55	—	56.97	0.38
SFT + PPO (Rlex + Rcons)2026.01		5.35	64.65	61.65	0.4
SFT2026.01		5.04	64.96	63.38	0.4
SFT + PPO (Rsent + Rcons)2026.01		4.75	65.25	63.11	0.4
SFT + PPO (Rsent + Rlex + Rcons)2026.01		4.75	65.25	62.69	0.4
GPT-4o-mini2026.01		4.59	65.41	59.64	0.4
SFT2026.01		1.89	51.89	61.91	0.38
SFT + PPO (Rlex + Rcons)2026.01		1.08	48.92	59.5	0.38
SFT + PPO (Rsent + Rlex + Rcons)2026.01		0.89	50.89	60.9	0.38
SFT + PPO (Rsent + Rcons)2026.01		0.7	50.7	61.27	0.38