Visual Question Answering on GeoHeight-Bench+ benchmark leaderboard · SOTA2 Research

Visual Question Answering on GeoHeight-Bench+

0Error Rate (ER)GPT-4o

Evaluation Results

Method	Links
GPT-4o2026.03		0	0.68	15.03	6.01	5.74	29.05	38.51	25.9	15.12
InternVL2-2B2026.03		0.68	0.56	14.18	7.07	6.41	15.52	37.33	13.51	11.91
InternVL2-8B2026.03		1.01	1.8	14.53	6.71	5.74	12.09	37.67	35.08	14.33
LLaVA-NeXT-7B-Mistral2026.03		2.27	0	14.86	6.07	4.62	16.63	38.01	11.82	11.79
Gemini-2.5-Flash2026.03		2.7	5.41	14.86	5.71	10.03	20.04	45.95	35.14	17.48
LLaVA1.5-7B2026.03		5.07	10.26	16.05	6.36	5.78	23.91	38.01	11.66	14.64
InternVL2-4B2026.03		5.41	16.58	10.14	7.42	8.2	20.29	38.34	33.61	17.5
LLaVA-NeXT-13B2026.03		7.09	12.62	15.54	6.71	5.93	12.99	38.07	12.44	13.92
LLaVA-NeXT-7B-Vicuna2026.03		7.43	10.59	16.39	7.77	5.06	18.08	37.82	11.88	14.38
Ministral3-8B2026.03		8.6	2.01	12.77	5.26	4.87	13.64	37.75	11.73	12.08
LLaVA1.5-13B2026.03		9.12	11.5	15.88	6.01	6.35	23.78	31.42	17.12	15.15
Pixtral-12B2026.03		10.75	2.96	14.6	5.34	5.29	15.11	37.6	12.98	13.08
LLaVA-NeXT-34B2026.03		11.32	11.38	15.24	7.27	6.2	19.76	37.91	14.26	15.42
LISA-7B (FT)2026.03		42.76	40.03	18.96	48.74	39.6	31.06	43.11	61.29	40.69
GeoHeightChat2026.03		76.35	52.62	18.58	97.53	94.87	48.48	55.41	80.91	65.59