Web-based Question Answering on WebWalkerQA Full Set
52.5Overall Success RateMANGO
Evaluation Results
| Method | Links | |
|---|---|---|
| MANGOModel=GPT-5-mini2026.04 | 52.5 | |
| MANGOModel=Qwen3-32B2026.04 | 28.38 | |
| MANGOModel=Qwen3-14B2026.04 | 26.03 | |
| WebWalkerModel=GPT-5-mini2026.04 | 25.74 | |
| MANGOModel=Qwen3-8B2026.04 | 21.91 | |
| AgentOccamModel=GPT-5-mini2026.04 | 20.29 | |
| MANGOModel=Qwen3-4B2026.04 | 17.06 | |
| WebWalkerModel=Qwen3-32B2026.04 | 16.76 | |
| WebWalkerModel=Qwen3-14B2026.04 | 14.71 | |
| WebWalkerModel=Qwen3-8B2026.04 | 12.65 | |
| WebWalkerModel=Qwen3-4B2026.04 | 12.5 | |
| AgentOccamModel=Qwen3-32B2026.04 | 10.59 | |
| AgentOccamModel=Qwen3-14B2026.04 | 6.32 | |
| AgentOccamModel=Qwen3-8B2026.04 | 4.85 | |
| AgentOccamModel=Qwen3-4B2026.04 | 4.56 |