Semantic Text Understanding on TED-6K 1.0 (test) benchmark leaderboard · SOTA2 Research

Semantic Text Understanding on TED-6K 1.0 (test)

57.42Norm-AvgGRAN-TED

Evaluation Results

Method	Links
GRAN-TED2025.12		57.42	—	—
Qwen3-VL-32B-Instruct2025.12		57.24	54.6	55.91
GRAN-TED w/o weighting2025.12		57.22	55.87	55.13
Qwen3-32B-Instruct2025.12		56.98	53.23	54.42
Qwen3-VL-8B-Instruct2025.12		56.81	55.37	54.25
Qwen3-VL-8B-Thinking2025.12		56.51	55.67	55.4
MiMo-VL-7B2025.12		56.48	54.54	56.29
Ovis-2.5-9B2025.12		56.26	54.66	54.92
Qwen3-8B-Base2025.12		55.94	53.37	54.19
Qwen3-8B-Instruct2025.12		55.77	53.62	53
Qwen3-4B-Instruct2025.12		55.25	53.64	53.58
Qwen3-VL-4B-Instruct2025.12		55.2	54.03	54.92
Qwen3-4B-Base2025.12		54.98	52.84	53.06
Qwen3-4B-Thinking2025.12		54.86	52.9	54.1
AltCLIP2025.12		51.86	47.15	46.59
UMT5-XXL2025.12		51.41	44.6	46.89