Fragment Retrieval on WeChat (test) benchmark leaderboard · SOTA2 Research

Fragment Retrieval on WeChat (test)

57.21PrecisionF2RVLM-Qwen2-VL-7B

Evaluation Results

Method	Links
F2RVLM-Qwen2-VL-7B2026.06		57.21	67.46	62.07	55.46
GPT-4o2026.06		56.11	41.82	48.89	42.85
Doubao-Seed-1.62026.06		55.22	28.47	38.63	34.06
F2RVLM-Qwen2.5-VL-7B2026.06		54.83	65.16	59.6	52.39
Gemini-2.5-Flash2026.06		51.89	49.8	53.21	48.22
Claude-Sonnet-42026.06		51.15	40.88	46.3	40.5
MiMo-7B-RL2026.06		49.46	48.22	49.05	41.84
MiMo-7B-SFT2026.06		49.14	46.58	47.98	40.71
DeepSeek-VL2-Small-16B2026.06		47.43	19.48	27.91	24.51
F2RVLM-Qwen2.5-VL-3B2026.06		45.24	70.86	55.6	48.09
Qwen2-VL-7B2026.06		42.8	69.6	53.54	45.75
MiMo-7B-RL2026.06		41.74	24.48	30.94	23.68
Ovis2-2B2026.06		40.83	54.51	46.69	33.16
Qwen2.5-VL-7B2026.06		39.99	71.3	51.71	43.65
InternVL3-8B2026.06		34.08	75.52	47.07	36.65
Qwen2.5-VL-3B2026.06		33.75	75.82	47.6	39.56
Qwen2.5-VL-72B2026.06		32.55	36.95	36.6	25.26
InternVL3-2B2026.06		31.59	73.57	44.32	33.34
E5-V-Embedding2026.06		30.33	47.13	36.99	24.4
DeepSeek-VL2-Tiny2026.06		29.83	21.12	24.73	16.79
GME-Embedding2026.06		29.63	53.11	38.22	26.68
F2RVLM-Qwen2-VL-2B2026.06		26.97	72	40.46	30.9
InternVL3-1B2026.06		23.88	79.86	37.42	28.33
Qwen2-VL-2B2026.06		23.2	76.03	36.65	26.55
LLaVA-1.5-7B-hf2026.06		20.91	81.27	33.61	22.99
CLIP-Embedding2026.06		20.44	49.91	31.56	17.62
mPLUG-Owl3-2B2026.06		18.35	45.99	26.37	11.61
BLIP2-Embedding2026.06		15.22	52.74	25.05	4.94
Qwen2.5-VL-7B2026.06		10.22	16.34	12.58	0