Plot-to-Code Generation on Plot2Code
71.95Text-Match AccuracyPPoT + Qwen2.5 VL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PPoT + Qwen2.5 VLModel Size=7B, k (LLM samples)=5, m (PPoT samples)=52026.04 | 71.95 | — | — | — | — | |
| Qwen2.5 VLModel Size=7B, k (LLM samples)=5, m (PPoT samples)=02026.04 | 69.92 | — | — | — | — | |
| Gemini-2.5-ProModel Category=Proprietary Models2026.04 | 69.1 | 88.6 | 7.45 | — | — | |
| PPoT + Qwen2.5 VLModel Size=3B, k (LLM samples)=5, m (PPoT samples)=52026.04 | 64.84 | — | — | — | — | |
| CharTide-8BModel Category=Open-Source Chart-Domain, Training Stage=RL2026.04 | 64.6 | 91.7 | 5.93 | — | — | |
| Qwen2.5 VLModel Size=3B, k (LLM samples)=5, m (PPoT samples)=02026.04 | 63.86 | — | — | — | — | |
| MM-ReCoder# Turns=12026.04 | 63.2 | 88.6 | — | — | 4.62 | |
| MM-ReCoder-Single# Turns=12026.04 | 62.8 | 89.4 | — | — | 5.1 | |
| MM-ReCoder# Turns=42026.04 | 62.7 | 98.5 | — | — | 5.24 | |
| MM-ReCoder# Turns=22026.04 | 62.2 | 97.7 | — | — | 5.17 | |
| GPT-5Model Category=Proprietary Models2026.04 | 61.9 | 87.8 | 7.28 | — | — | |
| Qwen3-VL-235B-A22B# Turns=12026.04 | 60.9 | 90.2 | — | — | 5.93 | |
| Qwen3-VL-235B-A22B# Turns=22026.04 | 60.4 | 91.7 | — | — | 5.79 | |
| GPT-4o (reproduced)# Turns=12026.04 | 59.8 | 82.6 | — | — | 5.42 | |
| CharTide-7BModel Category=Open-Source Chart-Domain, Training Stage=RL2026.04 | 59.6 | 89.4 | 5.6 | — | — | |
| GPT-4o (reproduced)# Turns=22026.04 | 58.4 | 90.2 | — | — | 6.12 | |
| CharTide-7B-SFTModel Category=Open-Source Chart-Domain, Training Stage=SFT2026.04 | 58.2 | 88.6 | 5.17 | — | — | |
| CharTide-8B-SFTModel Category=Open-Source Chart-Domain, Training Stage=SFT2026.04 | 58.1 | 86.4 | 5.46 | — | — | |
| GPT-4V2026.02 | 57.7 | 84.1 | 6.48 | — | — | |
| GPT-4V# Turns=12026.04 | 57.7 | 84.1 | — | 5.58 | — | |
| Claude-3-opus2026.02 | 57.5 | 84.1 | 4.37 | — | — | |
| Claude-3-opus# Turns=12026.04 | 57.5 | 84.1 | — | 3.8 | — | |
| GPT-4o2026.02 | 56.3 | 88.6 | 5.71 | — | — | |
| GPT-4o# Turns=12026.04 | 56.3 | 88.6 | — | 5.71 | — | |
| Qwen2.5-VL-72B# Turns=22026.04 | 55.4 | 87.1 | — | — | 4.96 | |
| ChartSpec + 4kParams=7B2026.02 | 55.3 | 88.7 | 6.32 | — | — | |
| Qwen2.5-VL-72B# Turns=12026.04 | 55 | 82.6 | — | — | 4.67 | |
| ChartCoderParams=7.0B2026.02 | 54.5 | 87.9 | 4.5 | — | — | |
| ChartCoder# Turns=12026.04 | 54.5 | 87.9 | — | 4.5 | 3.45 | |
| GeminiProVision2026.02 | 53.6 | 68.2 | 5.06 | — | — | |
| GeminiProVision# Turns=12026.04 | 53.6 | 68.2 | — | 3.69 | — | |
| ChartMaster-7BModel Category=Open-Source Chart-Domain2026.04 | 53.6 | 89.4 | 4.73 | — | — | |
| Qwen2-VL-72BParams=72.0B2026.02 | 53.4 | 72 | 4.26 | — | — | |
| Qwen2-VL-72B# Turns=12026.04 | 53.4 | 72 | — | 4.26 | — | |
| GPT-4oModel Category=Proprietary Models2026.04 | 52.6 | 87.1 | 5.66 | — | — | |
| ChartSpec + 3kParams=7B2026.02 | 50.3 | 86.4 | 6.28 | — | — | |
| VinciCoder-8BModel Category=Open-Source Chart-Domain2026.04 | 49.8 | 85.6 | 4.49 | — | — | |
| Qwen3-VL-8B# Turns=12026.04 | 49 | 73.5 | — | — | 4.11 | |
| Qwen2.5-VL-7B# Turns=12026.04 | 47.8 | 75 | — | — | 3.51 | |
| Qwen2.5-VL-7B(backbone)Params=7.0B2026.02 | 47.1 | 73.5 | 4.71 | — | — | |
| Qwen3-VL-8B# Turns=22026.04 | 46.8 | 84.1 | — | — | 4.31 | |
| Qwen3-VL-30B-A3BModel Category=Open-Source General-Domain2026.04 | 46.7 | 87.1 | 4.85 | — | — | |
| InternVL2-Llama3-76BParams=76.0B2026.02 | 46.6 | 85.6 | 3.89 | — | — | |
| InternVL2-Llama3-76B# Turns=12026.04 | 46.6 | 85.6 | — | 3.89 | — | |
| Qwen3-VL-235B-A22BModel Category=Open-Source General-Domain2026.04 | 46.2 | 84.8 | 5.19 | — | — | |
| TinyChartParams=3B2026.02 | 44.6 | 43.2 | 2.19 | — | — | |
| TinyChart# Turns=12026.04 | 44.6 | 43.2 | — | 2.19 | — | |
| Qwen2.5-VL-7B# Turns=22026.04 | 43.8 | 78.5 | — | — | 3.25 | |
| InternVL2-26BParams=26.0B2026.02 | 43.1 | 81.3 | 3.42 | — | — | |
| InternVL2-26B# Turns=12026.04 | 43.1 | 81.3 | — | 3.42 | — | |
| PPoT + Qwen2.5 VLModel Size=7B, k (LLM samples)=1, m (PPoT samples)=52026.04 | 42.48 | — | — | — | — | |
| PPoT + Qwen2.5 VLModel Size=3B, k (LLM samples)=1, m (PPoT samples)=52026.04 | 41.89 | — | — | — | — | |
| ChartLlamaParams=13B2026.02 | 40.3 | 58.4 | 2.32 | — | — | |
| ChartLlama# Turns=12026.04 | 40.3 | 58.4 | — | 2.32 | — | |
| Qwen2.5 VLModel Size=3B, k (LLM samples)=1, m (PPoT samples)=02026.04 | 39.44 | — | — | — | — | |
| LLaVA-Next-Mistral-7BParams=7.6B2026.02 | 38.7 | 72 | 3.06 | — | — | |
| LLaVA-Next-Mistral-7B# Turns=12026.04 | 38.7 | 72 | — | 2.87 | — | |
| MiniCPM-Llama3-V2.5Params=8.4B2026.02 | 37.3 | 76.3 | 2.61 | — | — | |
| MiniCPM-Llama3-V2.5# Turns=12026.04 | 37.3 | 76.3 | — | 2.61 | — | |
| InternVL2-8BParams=8.1B2026.02 | 37.1 | 77.3 | 2.78 | — | — | |
| InternVL2-8B# Turns=12026.04 | 37.1 | 77.3 | — | 2.78 | — | |
| Qwen2.5 VLModel Size=7B, k (LLM samples)=1, m (PPoT samples)=02026.04 | 36.7 | — | — | — | — | |
| Qwen3-VL-8BModel Category=Open-Source General-Domain2026.04 | 36.3 | 76.5 | 3.91 | — | — | |
| MSRL-7B-SFTModel Category=Open-Source Chart-Domain, Training Stage=SFT2026.04 | 34.7 | 77.3 | 3.71 | — | — | |
| Qwen2-VL-7BParams=7.0B2026.02 | 33.8 | 68.2 | 3.1 | — | — | |
| Qwen2-VL-7B# Turns=12026.04 | 33.8 | 68.2 | — | 3.1 | — | |
| Qwen2.5-VL-7BModel Category=Open-Source General-Domain2026.04 | 33.7 | 68.9 | 3.04 | — | — | |
| VinciCoder-7BModel Category=Open-Source Chart-Domain2026.04 | 33.6 | 68.9 | 3.39 | — | — | |
| Qwen2.5-VL-72BModel Category=Open-Source General-Domain2026.04 | 33.2 | 59.3 | 3.61 | — | — | |
| DeepSeek-VL-7BParams=7.3B2026.02 | 32.6 | 64.4 | 3.69 | — | — | |
| DeepSeek-VL-7B# Turns=12026.04 | 32.6 | 64.4 | — | 2.26 | — | |
| MSRL-7BModel Category=Open-Source Chart-Domain2026.04 | 31.9 | 62.9 | 3.24 | — | — | |
| ChartCoder-7BModel Category=Open-Source Chart-Domain2026.04 | 31.1 | 68.9 | 2.73 | — | — |