Author Response Generation on ARG Dataset

82GFP (Supervised)Llama-3.3

Evaluation results

Method
Llama-3.32026.01	82	14.2	3.8	—	—	—	82	100	82	—	—	—	—	82.9	51.3	51.7
DeepSeek2026.01	81.5	14.4	4.2	—	—	—	96	100	64	—	—	—	—	87.9	64.2	63.8
Llama-3.32026.01	78.8	15.7	5.5	—	—	—	82	100	84	61.9	47	49	72.8	80.4	46.7	50.4
GPT-4o2026.01	78.1	19.2	2.8	43.2	53.2	3.6	339	—	—	—	—	—	—	92.9	68.8	70.8
GPT-4o2026.01	77.4	19.8	2.8	—	—	—	158	85.4	11	—	—	—	—	86.7	63.3	63.3
Llama-3.32026.01	77.1	17.3	5.6	42	54.2	3.9	198	—	—	—	—	—	—	82.9	58.8	57.5
Llama-3.32026.01	77	18.3	4.7	—	—	—	214	—	—	48.6	70.5	53.3	82.5	85	57.5	59.2
Llama-3.32026.01	76.6	21.5	1.9	66.4	31.9	1.7	169	—	—	—	—	—	—	80	55	56.7
DeepSeek2026.01	76.2	20.3	3.5	—	—	—	93	100	63	62.6	58.2	56.3	77.9	86.7	58.8	62.5
GPT-4o2026.01	76.2	21.5	2.3	—	—	—	336	—	—	38.6	78.4	47.7	79.4	91.3	69.2	70
Llama-3.32026.01	76	21.7	2.3	53.4	42.6	4	183	—	—	—	—	—	—	85	60.8	60.8
DeepSeek2026.01	75.4	21.8	2.8	—	—	—	179	—	—	55.4	71	57.7	82.3	88.8	66.3	67.1
Phi-42026.01	74.8	20	5.2	—	—	—	343	45.8	-27	—	—	—	—	77.9	58.3	57.9
Qwen32026.01	74.4	21.4	4.2	49.6	46.3	4.1	205	—	—	—	—	—	—	91.3	72.1	71.7
GPT-4o2026.01	74.4	23.8	1.9	—	—	—	156	91.7	10	50.6	74.4	56.7	83.4	87.9	59.6	62.1
DeepSeek2026.01	73.8	23.2	3.1	45.2	51.4	3.5	194	—	—	—	—	—	—	90.4	69.2	70
Qwen32026.01	73.4	22.3	4.4	—	—	—	125	100	38	—	—	—	—	90.4	70	70
DeepSeek2026.01	73.4	23.8	2.8	—	—	—	194	—	—	58.5	70.9	58.5	86.1	92.5	74.6	74.2
DeepSeek2026.01	72.8	23.1	4.1	—	—	—	97	100	63	66.1	58.5	58.7	85.2	91.3	70.4	70.4
Qwen32026.01	72.4	22.4	5.2	—	—	—	130	95.8	33	49.8	69.6	53.4	84.2	91.3	70	70
DeepSeek2026.01	72	27.3	0.7	69.5	27.2	3.3	154	—	—	—	—	—	—	85	60	62.1
Qwen32026.01	71.9	25.2	2.8	—	—	—	216	—	—	42.9	79.3	52.2	82.6	93.8	72.5	72.5
GPT-4o2026.01	71.5	26.8	1.7	—	—	—	163	79.2	5	50.7	68.8	55.4	77.2	90	67.5	67.5
GPT-4o2026.01	70.8	28.2	1	57.1	40	2.9	311	—	—	—	—	—	—	81.7	62.9	62.9
Phi-42026.01	70.5	23.6	5.9	35.8	59.2	5	343	—	—	—	—	—	—	77.1	57.9	57.9
DeepSeek2026.01	70.2	27.9	1.9	58.4	37.4	4.1	172	—	—	—	—	—	—	81.7	60.8	61.7
GPT-4o2026.01	69.5	27.7	2.9	—	—	—	367	—	—	37.3	79	47	88	92.5	72.1	72.1
GPT-4o2026.01	68.9	28.4	2.7	66.8	30.1	3.1	265	—	—	—	—	—	—	82.1	60	62.5
Phi-42026.01	68	25.3	6.7	—	—	—	284	—	—	48.5	64.4	50.4	79.1	82.1	58.3	59.6
Phi-42026.01	67.3	26.3	6.4	—	—	—	284	25	-55	47.1	64.2	49.7	75.5	82.9	60	61.3
Llama-3.32026.01	65.7	26.1	8.2	—	—	—	125	87.5	52	58.9	54.5	50.8	71.8	89.2	66.7	63.8
Llama-3.32026.01	64.7	31.9	3.4	—	—	—	304	—	—	37.2	70.7	44.4	80.6	88.8	75	70
Qwen32026.01	64.3	31.7	4	57.2	40.4	2.5	164	—	—	—	—	—	—	87.5	68.3	68.3
Qwen32026.01	61.8	34.2	4	62.8	34.3	2.9	127	—	—	—	—	—	—	89.2	63.8	65.4
Qwen32026.01	58.6	38	3.5	—	—	—	290	—	—	38.5	75.2	45.9	84.7	98.3	84.2	80
Phi-42026.01	57.7	36.4	5.9	47	49.4	3.6	428	—	—	—	—	—	—	78.3	58.3	59.2
Qwen32026.01	57.6	37.3	5	—	—	—	142	89.6	21	50.6	67.8	54.4	80.7	93.8	77.1	75.8
Phi-42026.01	57.5	37.4	5.1	50.9	45	4.2	127	—	—	—	—	—	—	82.1	56.3	57.9
Phi-42026.01	49	44.3	6.8	—	—	—	368	—	—	36.8	69.1	43.4	72.9	92.9	71.3	72.5
Phi-42026.01	48.9	44.2	6.9	—	—	—	312	10.4	-128	38.7	67.1	44.4	79	92.9	73.3	72.5
Llama-3.32026.01	48.3	41.4	10.3	5.4	91.2	3.4	126	—	—	—	—	—	—	76.3	39.6	43.8
Human2026.01	45.8	45.3	8.9	20	75.6	4.4	115	—	—	—	—	—	—	78.8	57.5	57.5
GPT-4o2026.01	44.3	46.7	9	3.3	92.8	3.9	247	—	—	—	—	—	—	82.5	47.9	54.7
DeepSeek2026.01	41.2	49.1	9.7	4.6	91.3	4.1	113	—	—	—	—	—	—	77.1	43.3	49.6
Phi-42026.01	36.2	54.2	9.6	30	92.6	4.4	161	—	—	—	—	—	—	77.5	44.6	48.3
Qwen32026.01	30.8	56.6	12.6	2.4	93	4.6	123	—	—	—	—	—	—	80.8	50.8	55.4