パターンを読み込んでいます…
⌨️
Terminal-Bench(TB)
Hard command-line tasks in isolated environments, each with a human-written solution and tests that decide whether the agent actually finished.
複雑さ: high評価とモニタリング
30秒でわかる概要
- 概要
- Poses hard command line tasks in isolated containers, each carrying a human-written solution and tests that decide afterwards whether the work was actually finished.
- 使いどころ
- You are evaluating agents that do infrastructure, build or debugging work in a shell, and you need correctness decided rather than judged.
- 注意点
- The score names a model and a harness together: the same weights under a weaker scaffold give up exactly where a stronger one reads the error and recovers.
このパターンについてAIエキスパートに質問
質問が入力済みの状態でアシスタントが開きます。送信前に内容を確認できます。
Terminal-Bench: 概要
Hard command-line tasks in isolated environments, each with a human-written solution and tests that decide whether the agent actually finished.
- One container per task, so runs cannot contaminate each other
- Every task carries a human-written reference solution
- Verified by tests rather than by inspecting the transcript
- Drawn from real infrastructure, data and debugging work
- Deliberately hard: frontier systems remain well short of solving it
- Harness-agnostic, so the agent and the scaffold are measured together
The Agent Architect
1つのパターン、1つのトレードオフ、1つの本番障害事例。エージェントシステムを構築する人のための短い週刊ブリーフィング。
週1回のメール、ワンクリックで購読解除できます。アドレスはブリーフィングの送信のみに使用します。
別名： CLI agent benchmark, Command line agent evaluation
参考文献
このパターンの根拠となる論文、仕様、リポジトリです。