Loading...
🏢
TheAgentCompany Benchmark(TAC)
Benchmarks LLM agents on consequential real-world tasks that would typically be completed by multiple job roles in a software engineering company.
Complexity: highEvaluation and Monitoring
In 30 seconds
- What
- Measures agent performance on multi-step professional tasks spanning software engineering roles, scoring full completion and partial credit separately.
- When to use
- Comparing LLM agents on realistic work scenarios before deployment, or tracking capability improvements across model versions.
- Watch out
- Partial credit scores can mask agents that fail at critical steps; a 39% partial score may hide zero-value incomplete work.
Loading technique guide…