Share your thoughts, 1 month free Claude Pro on usSee more

Chart-to-code generation on Plot2Code Chart2Python (test)

61.4Error Rate (ER)

InternVL3.5-2B

Updated 1mo ago

Evaluation Results

Method	Links
InternVL3.5-2B 2026.04		61.4	55.7	34.2	44.2
InternVL3.5-4B 2026.04		62.1	58.8	38	42.7
Qwen3-VL-2B 2026.04		68.9	64.2	41.2	50.1
LLaVA-v1.6-7B 2026.04		69.7	49.8	26.5	32.5
ChartMoE-8B 2026.04		70.5	58.9	37.6	26.9
DeepSeek-VL2-3B 2026.04		72.5	59.8	45.4	44.2
Phi-3.5-vision-4B 2026.04		72.7	63.8	43.1	42.6
InternVL3.5-8B 2026.04		74.2	61	47.2	49.1
Qwen3-VL-4B 2026.04		77.3	66.4	54.4	55.4
Qwen3-VL-8B 2026.04		78.8	68.1	57.3	56.9
ChartLlama-13B 2026.04		81.8	50.1	18.9	22.4
CharLuMA-1.3B 2026.04		83.3	64.3	50.6	47.2
Claude-Haiku-3.5 2026.04		87.1	72.8	60.6	56.8
ChartCoder-7B 2026.04		87.9	65.7	58.2	56.6
GPT-4o-mini 2026.04		90.2	77.8	63.8	67
GPT-4o 2026.04		92.4	83.6	76.7	75.4
GPT-5-mini 2026.04		93.2	85.9	72.7	72.8
Claude-Sonnet-4 2026.04		95.5	81.2	81	76.8
CharLuMA-6.7B 2026.04		96.2	68.3	62.2	60.5