Web Agent on WorkArena L2
4.7Success RateWEASEL (10K steps)
Evaluation Results
| Method | Links | |
|---|---|---|
| WEASEL (10K steps)Model=Qwen2.5-7B-Instruct, # Data=10K, # Epochs=4, Time=12.0 hr, Speed-up=11.3×2026.05 | 4.7 | |
| WEASEL (10K steps)Model=Qwen3-8B, # Data=10K, # Epochs=2, Time=8.3 hr, Speed-up=10.7×2026.05 | 4.3 | |
| Pruning + LLM-Judge (10K steps)Model=Qwen3-8B, # Data=10K, # Epochs=2, Time=7.0 hr, Speed-up=12.6×2026.05 | 3.8 | |
| Pruning + Sampling (10K steps)Model=Qwen3-8B, # Data=10K, # Epochs=2, Time=7.0 hr, Speed-up=12.6×2026.05 | 3.4 | |
| Pruning + Sampling (10K steps)Model=Qwen2.5-7B-Instruct, # Data=10K, # Epochs=4, Time=12.0 hr, Speed-up=11.3×2026.05 | 3 | |
| Pruning + LLM-Judge (10K steps)Model=Qwen2.5-7B-Instruct, # Data=10K, # Epochs=4, Time=12.0 hr, Speed-up=11.3×2026.05 | 3 | |
| WEASEL (10K steps)Model=Gemma3-4B-IT, # Data=10K, # Epochs=2, Time=6.4 hr, Speed-up=12.5×2026.05 | 3 | |
| Pruning (52K steps)Model=Qwen3-8B, # Data=52K, # Epochs=2, Time=44.0 hr, Speed-up=2.0×2026.05 | 2.6 | |
| Pruning + Sampling (10K steps)Model=Gemma3-4B-IT, # Data=10K, # Epochs=2, Time=6.4 hr, Speed-up=12.5×2026.05 | 2.1 | |
| Pruning + LLM-Judge (10K steps)Model=Gemma3-4B-IT, # Data=10K, # Epochs=2, Time=6.4 hr, Speed-up=12.5×2026.05 | 2.1 | |
| Full (52K steps)Model=Qwen3-8B, # Data=52K, # Epochs=2, Time=88.5 hr, Speed-up=1.0×2026.05 | 2.1 | |
| Qwen3-8BModel=Qwen3-8B2026.05 | 1.7 | |
| Full (52K steps)Model=Qwen2.5-7B-Instruct, # Data=52K, # Epochs=4, Time=136.0 hr, Speed-up=1.0×2026.05 | 0.4 | |
| Pruning (52K steps)Model=Qwen2.5-7B-Instruct, # Data=52K, # Epochs=4, Time=62.0 hr, Speed-up=2.2×2026.05 | 0.4 | |
| Qwen2.5-7B-InstructModel=Qwen2.5-7B-Instruct2026.05 | 0 | |
| Gemma3-4B-ITModel=Gemma3-4B-IT2026.05 | 0 | |
| Full (52K steps)Model=Gemma3-4B-IT, # Data=52K, # Epochs=2, Time=80.0 hr, Speed-up=1.0×2026.05 | 0 | |
| Pruning (52K steps)Model=Gemma3-4B-IT, # Data=52K, # Epochs=2, Time=30.0 hr, Speed-up=2.7×2026.05 | 0 |