Red-Teaming Auto Mode: Improving Blocking Classifiers Against Malign Coding Agents
Alex Remedios, Simon Storf, Fabien Roger, John Hughes
実装難易度
Hard
推論・学習コスト
High
想定用途
分類
概要
To keep coding agents from going off the rails, production systems now review each proposed action with a blocking monitor that can reject it before it runs (Auto Mode in Claude Code, Guardian in OpenAI's Codex). Prior evaluations of such monitors largely measure robustness to accidental harm or prompt injections from untrusted sources looking to hijack the agent. Less understood is how they hold…
何が新しいか
To keep coding agents from going off the rails, production systems now review each proposed action with a blocking monitor that can reject it before it runs (Auto Mode in Claude Code, Guardian in…
何に使えるか
分類
実装情報
- Paper URL
- あり
- GitHub URL
- あり
実装チェックリスト
実装または配布ページ
OKコードまたはモデル配布ページから検証を始められます。
一次情報リンク
OKPaper / GitHub
検証しやすさ
要確認大規模モデル、生成、強化学習など再現コストが高い要素を含みます。
計算資源
要確認学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
ライセンス
未取得配布元のLICENSE、モデルカード、Paperの利用条件を確認してください。
商用利用
未取得研究利用限定、データセット由来制限、API規約の有無を確認してください。
自社データで試すなら
製造業・材料開発のExcel/CSVデータに落とし込むための最初の手順です。
- 1まず自社データを、入力条件、目的変数、評価したい指標に分けて整理します。
- 2LightGBMやRandom Forestなどのベースラインを先に作り、この手法と比較します。
- 3評価指標はR2/RMSE、AUC、異常検知の再現率、実験回数削減率など、現場の意思決定に近いものを選びます。
- 4SHAPや特徴量重要度で、効いている因子が物理・化学・工程知識と矛盾しないか確認します。
実装難易度
Hard - 大規模モデル、生成、強化学習など再現コストが高い要素を含みます。
必要リソース
- GPU目安: High
- データセット: 論文・リポジトリ側の指定を確認してください。
- 学習要否: 再学習や評価環境の準備が必要になる可能性があります。
- 学習や高解像度推論ではGPUメモリと実行時間に注意が必要です。
実務で使う場合の注意点
- ライセンスと商用利用条件は、Paper / GitHub / Hugging Face の配布元で確認してください。
- 精度、再現性、計算コストはデータセットや評価条件に依存します。
- 個人情報や機密データを扱う場合は、入力データの保存先と外部API利用条件を確認してください。
関連記事
paperless-ngx — A community-supported supercharged document management system: scan, index and archive all your documents
paperless-ngxは、コミュニティによってサポートされたスーパーチャージドのドキュメント管理システムで、ドキュメントのスキャン・インデックス・アーカイブが可能である。
SAFARI: An Industrial Benchmark for LLM-Assisted Hazard Analysis and Risk Assessment
Large language models (LLMs) are increasingly considered for safety-critical engineering, yet their reliabilit
AgriScope: Pixel-Grounded Multimodal Understanding for Agricultural Images
Agricultural image understanding requires fine-grained recognition of plant diseases, pests, crop structures,
Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations
Safety evaluations for large language models rely on surface-form classifiers that report declining harm score