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

Chart-based Reasoning on CharXivRQ

67.9Accuracy

Gemini 2.5 Pro

Updated 9d ago

Evaluation Results

Method	Links
Gemini 2.5 Pro 2026.04		67.9
Claude-3.7-Sonnet 2026.02		64.2
Octopus-8B (Ours) 2026.02		55.7
Qwen3-VL-8B-Instruct + GSPO 2026.02		55.3
OpenAI-o1 2026.02		55.1
MiMo-VL-7B-SFT 2026.02		54.8
MiMo-VL-7B-RL 2026.02		53.2
Qwen3-VL-8B-Thinking 2026.02		53
OpenVLThinkerV2 2026.04		53
Qwen3-VL-8B-Instruct + DAPO 2026.02		52.8
Qwen3-VL-8B-Instruct + SRPO 2026.02		52.7
Qwen3-VL GDPO 2026.04		51.6
Qwen3-VL-8B-Instruct + GRPO 2026.02		51.4
Qwen3-VL-8B-Instruct + SRPO 2026.02		51.2
Qwen3-VL-8B-Instruct + GRPO 2026.02		50.7
GPT-4o 2026.02		50.5
Qwen3-VL GRPO 2026.04		50.5
Qwen3-VL-8B-Instruct + GSPO 2026.02		47.9
GPT-4o 2026.04		47.1
ARES-7B 2026.04		47
Qwen3-VL-8B-Instruct 2026.02		45.1
OVR-7B 2026.04		44.5
Qwen3-VL-Instruct-8B 2026.04		44.5
InternVL3.5-8B-RL 2026.02		44.4
OneThinker-8B 2026.04		44
Vision-G1 2026.04		41
VL-Rethinker-7B 2026.04		39.8
MM-Eureka-7B 2026.04		39.5
OpenVLThinker-7B 2026.04		39.3