Context-aware turn-taking on AMI benchmark leaderboard · SOTA2 Research

Context-aware turn-taking on AMI

92.41I1 ScoreLLaMA3.1-8b-instruct

Evaluation Results

Method	Links
LLaMA3.1-8b-instruct2026.03		92.41	69.86	30.57	9.55	50.72	47.09	50.35
LLaMA3.1-8b-instruct2026.03		92.41	69.86	30.57	9.55	50.72	47.09	50.35
Qwen3-4b-instruct2026.03		92.41	77.63	39.81	7.64	56.68	53.53	56.32
Qwen3-4b-instruct2026.03		92.41	77.63	39.81	7.64	56.68	53.53	56.32
Qwen3-8b2026.03		90.51	68.95	48.82	6.37	56.26	54.53	55.99
Qwen3-8b2026.03		90.51	68.95	48.82	6.37	56.26	54.53	55.99
Mistral-7b-instruct2026.03		89.24	83.33	14.69	8.28	49.45	41.59	48.93
Mistral-7b-instruct2026.03		89.24	83.33	14.69	8.28	49.45	41.59	48.93
Qwen2.5-7b2026.03		88.61	70.09	28.2	19.11	50.72	47.34	50.37
Qwen2.5-7b2026.03		88.61	70.09	28.2	19.11	50.72	47.34	50.37
gpt-5.22026.03		75.95	55.71	66.59	32.48	59.23	59.2	59.21
gpt-5.22026.03		75.95	55.71	66.59	32.48	59.23	59.2	59.21
gpt-oss-20b2026.03		74.05	45.89	65.4	31.21	54.72	54.72	54.74
gpt-oss-20b2026.03		74.05	45.89	65.4	31.21	54.72	54.72	54.74
gemini-3.1-pro2026.03		52.53	39.27	88.63	54.14	60.77	59.55	61.03
gemini-3.1-pro2026.03		52.53	39.27	88.63	54.14	60.77	59.55	61.03