Multi-user Dialogue Response Generation on PAT benchmark leaderboard · SOTA2 Research

Multi-user Dialogue Response Generation on PAT

46.3P-CovLLaMA-70B

Evaluation Results

Method	Links
LLaMA-70B2026.04		46.3	42.5	80.6	73.6	65.6	58.6	51.2	26.7	36.4	78
LLaMA-70B2026.04		43.5	40.6	80.5	73.4	65.1	58	49.7	24.5	37.4	77.7
Anthropic Claude2026.04		42.4	41	74.3	66.8	57	48.1	40.1	10.5	23.6	79.2
LLaMA-70B2026.04		42	39.5	78.2	71.3	59.5	49.5	41.5	11	22.9	89.4
Anthropic Claude2026.04		40.8	36.3	76.4	68.1	57.8	48.7	37.4	10.3	22.8	81.7
Anthropic Claude2026.04		39.4	30.4	76.9	68.7	58.4	49.3	39	10.2	22.9	81
OpenAI GPT-4o2026.04		37.4	30.6	79.8	70.5	59.5	50.2	33.5	9.7	21.7	83.1
Google Gemini2026.04		37	35.2	60.8	54.5	46.9	40.5	29.9	9.8	20	85.1
OpenAI GPT-3.52026.04		35.3	36.3	59	52.7	44.8	38.8	25.6	7.5	16.8	86.7
Google Gemini2026.04		35.2	39.2	52.8	47.5	40.6	34.8	27.3	8.4	17.5	86.3
OpenAI GPT-3.52026.04		33.3	30.5	54.6	48.9	41.6	35.5	24.2	7.4	15.9	88.1
Google Gemini2026.04		33.1	29.2	57.4	51.4	43.8	37.6	27.7	9.3	18.4	85.4
OpenAI GPT-4o2026.04		30.5	28.4	77.7	68.6	58.1	49.2	32.4	9.9	21.3	83.9
OpenAI GPT-3.52026.04		30.1	37.9	49.7	44.3	37.7	32.2	23.3	6.8	15.3	89
OpenAI GPT-4o2026.04		30.1	29.4	74.8	66.3	56.2	47.8	32.6	9.7	20.7	84.7