Fragment Retrieval on MLDR (val) benchmark leaderboard · SOTA2 Research

Fragment Retrieval on MLDR (val)

84.02PrecisionF2RVLM-Qwen2-VL-7B

Evaluation Results

Method	Links
F2RVLM-Qwen2-VL-7B2026.06		84.02	90.67	87.25	80.6
F2RVLM-Qwen2.5-VL-7B2026.06		83.24	91.6	87.24	80.55
Qwen2-VL-7B2026.06		83.15	90.72	86.81	79.92
F2RVLM-Qwen2.5-VL-3B2026.06		82.74	91.65	87	80.19
MiMo-7B-SFT2026.06		81.61	93.25	87.08	80.35
Qwen2.5-VL-7B2026.06		81.52	91.42	86.23	79
MiMo-7B-RL2026.06		80.9	93.46	86.71	79.78
InternVL3-8B2026.06		80.67	92.98	86.43	79.35
Qwen2.5-VL-3B2026.06		80.35	91.41	85.57	77.98
F2RVLM-Qwen2-VL-2B2026.06		79.64	89.55	84.35	76.07
DeepSeek-VL2-Small-16B2026.06		77.52	91.13	83.82	75.23
Ovis2-2B2026.06		76.97	42.48	54.82	43.98
InternVL3-2B2026.06		75.73	91.86	83.07	74.11
Qwen2-VL-2B2026.06		74.97	92.99	83.07	74.18
mPLUG-Owl3-2B2026.06		74.86	83.18	78.88	67.59
Doubao-Seed-1.62026.06		73.83	42.19	54.67	42.57
GPT-4o2026.06		70.43	52.49	60.32	45.91
Gemini-2.5-Flash2026.06		70.18	69.3	69.87	54.66
LLaVA-1.5-7B-hf2026.06		68.15	92.06	78.43	66.95
MiMo-7B-RL2026.06		67.68	55.19	61.3	45.52
Claude-Sonnet-42026.06		67.21	58.09	62.8	46.57
InternVL3-1B2026.06		67.1	89.89	77	64.64
DeepSeek-VL2-Tiny2026.06		62.97	84.25	72.23	56.59
GME-Embedding2026.06		62.27	23.75	35.52	24.17
Qwen2.5-VL-72B2026.06		61.11	67.14	64.09	44.61
E5-V-Embedding2026.06		53.85	48.83	51.82	28.92
CLIP-Embedding2026.06		48.74	30.8	42.73	21.14
BLIP2-Embedding2026.06		31.88	2.96	23.47	0
Qwen2.5-VL-7B2026.06		21.2	4.27	7.84	0