Vision-Language Reasoning on DisasterBench (test)

66.67RAEGPT-5

Evaluation results

Method
GPT-52026.06	66.67	84.06	92.51	78.57	74.56	87.06	80.09	97.48	90.11	84.83	644
InternVL2-8B2026.06	61.4	63.77	81.43	65.18	42.6	53.87	72.4	92.44	84.98	68.23	255
Gemini-2.5-pro2026.06	58.77	84.06	94.14	76.49	70.41	87.06	80.32	93.28	90.49	84.17	1,323
Qwen2.5-VL-7B2026.06	56.14	67.03	85.02	72.32	40.83	63.29	70.14	92.44	86.69	71.6	425
Gemini-2.5-flash2026.06	55.26	83.33	92.83	63.1	69.82	77.64	71.04	91.6	87.07	78.03	1,321
Grok-42026.06	53.51	76.09	89.58	76.19	62.72	81.15	69.68	89.92	82.51	77.8	733
Claude-Sonnet-4-52026.06	52.63	80.07	91.86	68.45	71.01	82.28	76.02	93.28	86.12	79.93	457
InternVL-3-5-4B2026.06	51.75	55.8	82.08	57.14	59.76	72.86	54.3	85.71	78.14	67.63	459
DisasterVL2026.06	50.88	57.97	91.86	80.95	51.48	70.04	59.05	85.71	87.07	72.6	168
Gemini-2.0-flash2026.06	47.37	76.45	89.9	75.89	65.68	71.17	81.45	88.24	92.4	78.8	435
Qwen2.5-VL-3B2026.06	47.37	55.43	75.57	62.2	34.32	42.33	63.8	89.08	75.67	59.77	428
InternVL-3-5-8B2026.06	43.86	54.35	83.71	61.31	55.62	68.64	51.36	88.24	78.9	66.4	322
GPT-4o2026.06	40.35	76.81	86.32	58.33	53.85	76.37	71.95	83.19	79.85	73	335
Deepseek-VL2-S2026.06	31.58	54.71	92.51	61.31	53.25	64.56	61.54	88.24	79.47	67.37	359
Phi-3-Vision(4.15B)2026.06	31.58	59.42	83.71	74.7	31.95	44.59	67.19	84.03	83.84	63.9	272
InternVL-3-5-2B2026.06	30.7	45.29	82.08	50.89	41.42	60.76	46.83	67.23	73.76	58.67	355
GPT-4o mini2026.06	27.19	64.86	81.76	63.39	55.03	55.27	62.44	83.19	73.19	64	423
Janus-Pro-1B2026.06	14.91	43.84	76.55	53.57	28.99	57.52	34.62	60.5	61.41	51.97	354
Qwen2-VL-2B2026.06	13.16	42.75	69.06	52.98	24.26	22.78	44.57	69.75	58.94	43.87	352
Deepseek-VL2-T2026.06	13.16	38.41	78.5	40.18	36.69	54.85	39.37	46.22	39.73	46.23	190
InternVL-3-5-1B2026.06	12.28	33.7	80.13	38.99	40.83	49.65	31.22	48.74	50.57	45.6	380
InternVL2-2B2026.06	9.65	52.17	77.52	46.43	29.59	33.33	50.68	76.47	63.12	49.43	425