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

Context-aware turn-taking on SPGI

96.86I1 Score

LLaMA3.1-8b-instruct

Updated 1mo ago

Evaluation Results

Method	Links
LLaMA3.1-8b-instruct 2026.03		96.86	65.32	20.75	4.58	44.36	37.2	44.76
LLaMA3.1-8b-instruct 2026.03		96.86	65.32	20.75	4.58	44.36	37.2	44.76
Qwen3-4b-instruct 2026.03		94.94	55.07	16.99	10.34	42.25	36.81	42.6
Qwen3-4b-instruct 2026.03		94.94	55.07	16.99	10.34	42.25	36.81	42.6
Qwen3-8b 2026.03		94.94	43.7	33.7	11.28	42.46	39.31	42.73
Qwen3-8b 2026.03		94.94	43.7	33.7	11.28	42.46	39.31	42.73
Qwen2.5-7b 2026.03		89.28	67.95	25.72	16.3	48.15	43.68	48.48
Qwen2.5-7b 2026.03		89.28	67.95	25.72	16.3	48.15	43.68	48.48
gpt-5.2 2026.03		85.3	31.54	75.48	49.45	56.96	56.93	56.94
gpt-5.2 2026.03		85.3	31.54	75.48	49.45	56.96	56.93	56.94
Mistral-7b-instruct 2026.03		84.6	72.03	24.03	19.52	49.01	44.8	49.33
Mistral-7b-instruct 2026.03		84.6	72.03	24.03	19.52	49.01	44.8	49.33
gemini-3.1-pro 2026.03		79.35	30.31	91.32	69.07	64.57	63.93	64.45
gemini-3.1-pro 2026.03		79.35	30.31	91.32	69.07	64.57	63.93	64.45
gpt-oss-20b 2026.03		68.63	22.84	70.18	48.35	49.62	49.35	49.54
gpt-oss-20b 2026.03		68.63	22.84	70.18	48.35	49.62	49.35	49.54