Agentic Task Completion on AppWorld Easy normal (test)
86AccuracyACON UTCO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ACON UTCOAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 86 | 7.09 | 2.84 | |
| FIFOAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 84.2 | 5.85 | 2.89 | |
| ACON UTCOAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=Observation Compression2025.10 | 82.5 | 5.66 | 2.63 | |
| No compressionAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=None2025.10 | 80.7 | 7.57 | 2.98 | |
| ACON UTAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 77.2 | 6.45 | 2.43 | |
| ACON UTAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=Observation Compression2025.10 | 70.2 | 5.87 | 2.21 | |
| LLMLinguaAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 66.7 | 6.38 | 3.04 | |
| PromptingAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 66.7 | 6.36 | 2.84 | |
| PromptingAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=Observation Compression2025.10 | 64.9 | 4.92 | 1.88 | |
| RetrievalAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=History Compression2025.10 | 61.4 | 7.4 | 3.97 | |
| LLMLinguaAgent=gpt-4.1, Compressor=gpt-4.1, Compression Type=Observation Compression2025.10 | 54.4 | 5.78 | 2.33 |