Web Navigation QA on WebVoyager
4.55Average Action CountAgentOccam
Evaluation Results
| Method | Links | |
|---|---|---|
| AgentOccamModel=Qwen3, Backbone size=14B2026.04 | 4.55 | |
| AgentOccamModel=Qwen3, Backbone size=4B2026.04 | 5.34 | |
| MANGOModel=Qwen3, Backbone size=32B2026.04 | 5.35 | |
| AgentOccamModel=Qwen3, Backbone size=32B2026.04 | 6.07 | |
| MANGOModel=Qwen3, Backbone size=4B2026.04 | 6.21 | |
| MANGOModel=Qwen3, Backbone size=14B2026.04 | 6.26 | |
| AgentOccamModel=Qwen3, Backbone size=8B2026.04 | 6.48 | |
| WebWalkerModel=Qwen3, Backbone size=8B2026.04 | 7 | |
| WebWalkerModel=GPT-5 mini2026.04 | 7.38 | |
| WebWalkerModel=Qwen3, Backbone size=14B2026.04 | 7.44 | |
| MANGOModel=Qwen3, Backbone size=8B2026.04 | 7.83 | |
| WebWalkerModel=Qwen3, Backbone size=4B2026.04 | 9.11 | |
| WebWalkerModel=Qwen3, Backbone size=32B2026.04 | 9.32 | |
| AgentOccamModel=GPT-5 mini2026.04 | 9.46 | |
| MANGOModel=GPT-5 mini2026.04 | 14.18 |