Share your thoughts, 1 month free Claude Pro on usSee more

Controllable Writing on WritingBench (WB)

79.8WB-A Score

Qwen3-32B

Updated 4d ago

Evaluation Results

Method	Links
Qwen3-32B 2026.04		79.8	78.2	80.8	78.9	82.9	81.9
Claude 3.7 2026.04		78.2	77.9	76.5	79.3	79.2	80.8
DeepWriter-8B + RL-TCER 2026.04		76.5	75.3	74.8	74.2	76.7	75.2
DeepWriter-8B + RL-EndoR 2026.04		76	73.6	73.9	69.6	77.3	74.5
GPT-4o 2026.04		74.4	73.4	74.3	77.9	75.8	78
Qwen2.5-7B-Instruct + SFT + RL-TCER 2026.04		72.7	73.1	70.8	67.8	73.4	71.1
Qwen2.5-7B-Instruct + SFT + RL-EndoR 2026.04		72.6	70.9	70.1	64.6	73	68.7
DeepWriter-8B 2026.04		72.2	71.8	69.8	70.6	73.7	72.3
Qwen2.5-7B-Instruct + SFT 2026.04		69.7	69.8	67.9	63.1	71.3	66.3
Qwen3-8B 2026.04		68.7	68.9	67	67.2	71.2	71.3
LongWriter-8B + RL-TCER 2026.04		65.8	62.9	61.1	57.1	60.6	59.8
LongWriter-8B + RL-EndoR 2026.04		64.4	62.6	60.2	56.7	59.9	58.8
Claude 3.5 2026.04		59.5	57.6	56.3	59.3	62	67.7
Qwen2.5-7B-Instruct 2026.04		58.9	56.8	56.5	54	59.9	60
LongWriter-8B 2026.04		57.9	53.9	49	52	52.9	52
Qwen2.5-32B-Instruct 2026.04		52.5	49.8	51	49.6	53.9	54.2
Llama3.1-8B-Instruct 2026.04		47.5	45.6	43.7	42.3	48.6	50.8