Geometry Reasoning on Geometry3K (val)
44.93AccuracyQwen2.5-VL-3B-Instruct + VEPO
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL-3B-Instruct + VEPOBase Model Size=3B, Fine-tuning=VEPO2026.06 | 44.93 | |
| Qwen2.5-VL-3B-Instruct + GRPOBase Model Size=3B, Fine-tuning=GRPO2026.06 | 43.43 | |
| PAPO-DAPO-3BBase Model Size=3B2026.06 | 42.76 | |
| NoisyRollout-3BBase Model Size=3B2026.06 | 42.26 | |
| VPPO-3BBase Model Size=3B2026.06 | 42.1 | |
| Qwen2.5-VL-3B-Instruct + GRPO (Top 40% high entropy)Base Model Size=3B, Fine-tuning=GRPO, Constraint=Top 40% high entropy2026.06 | 40.77 | |
| R1-ShareVLBase Model Size=3B2026.06 | 39.93 | |
| Qwen2.5-VL-3B-Instruct + GRPO (Top 20% high entropy)Base Model Size=3B, Fine-tuning=GRPO, Constraint=Top 20% high entropy2026.06 | 38.28 | |
| PEPO_DBackbone=InternVL3-2B-Instruct2026.03 | 28.12 | |
| Qwen2.5-VL-3B-InstructBase Model Size=3B2026.06 | 26.96 | |
| PEPO_GBackbone=InternVL3-2B-Instruct2026.03 | 25.84 | |
| DAPOBackbone=Qwen2.5-VL-3B-Instruct2026.03 | 22.63 | |
| PEPO_DBackbone=Qwen2.5-VL-3B-Instruct2026.03 | 22.38 | |
| GRPOBackbone=InternVL3-2B-Instruct2026.03 | 22.08 | |
| PEPO_GBackbone=Qwen2.5-VL-3B-Instruct2026.03 | 21.91 | |
| DAPOBackbone=InternVL3-2B-Instruct2026.03 | 20.94 | |
| GRPOBackbone=Qwen2.5-VL-3B-Instruct2026.03 | 19 | |
| High-Entropy RLBackbone=Qwen2.5-VL-3B-Instruct2026.03 | 15.58 | |
| BaseBackbone=Qwen2.5-VL-3B-Instruct, Mode=zero-shot2026.03 | 11.58 | |
| High-Entropy RLBackbone=InternVL3-2B-Instruct2026.03 | 10.89 | |
| BaseBackbone=InternVL3-2B-Instruct, Mode=zero-shot2026.03 | 7.88 |