Analyst-style Data Analysis on 50k Agent Trace Suite
50Success CountDuckDB-WASM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DuckDB-WASMStack version=1.33.1-dev45.0, Agent Model=Anthropic Haiku, Trials per task=52026.05 | 50 | 2 | 0.203 | |
| SquirrelingStack version=0.12.16, Agent Model=Anthropic Haiku, Trials per task=52026.05 | 50 | 2 | 0.067 |