Reliable to Expressive: A Curriculum for Rubric-Following Safety Judges
Yongtaek Lim, Hyeji Choi, Minwoo Kim
実装難易度
Hard
推論・学習コスト
High
想定用途
分類
概要
Safety judges are increasingly deployed to evaluate model outputs against evolving criteria, yet recent meta-evaluation work shows they remain brittle under prompt and rubric variation, with false negative-rate swings of up to 0.24 reported for stylistic perturbations alone. We argue that safety judgment is fundamentally a rubric-following problem: a robust judge must apply the given evaluation…
何が新しいか
Safety judges are increasingly deployed to evaluate model outputs against evolving criteria, yet recent meta-evaluation work shows they remain brittle under prompt and rubric variation, with false…
何に使えるか
分類
実装情報
- Paper URL
- あり
実装チェックリスト
実装または配布ページ
要確認Paper onlyの可能性があるため再実装前提で確認してください。
一次情報リンク
OKPaper
検証しやすさ
要確認公式実装が見つからないため、論文から再実装する前提です。
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 - 公式実装が見つからないため、論文から再実装する前提です。
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
MUDIDI: A Two-Stage Framework for Multilingual Dictionary Digitization with Language Models
この研究では、低リソース言語や絶滅言語の辞書のデジタル化が重要であるが、マルチモーダル辞書をデジタル化する方法は今まで難しかったが、この研究では、最近のビジョン言語モデルを用いて辞書のデジタル化が容易になり、辞書内の文字
LLM-Guided Evolution for Medical Decision Pipelines
Adapting large language models (LLMs) to clinical workflows often requires costly fine-tuning or manual prompt
llm-ie — A comprehensive toolkit that provides building blocks for LLM-based named entity recognition, attribute extraction, and relation extraction pipelines.
A comprehensive toolkit that provides building blocks for LLM-based named entity recognition, attribute extrac
What the Eyes See, the LLMs Miss: Exploiting Human Perception for Adversarial Text Attacks
大規模言語モデル(LLM)を運用するコンテンツモデレーションシステムは、有害なオンラインコンテンツを防止するために重要な役割を果たします。しかし、これらのシステムの主な目標は単にトークナイズされたテキストを操作することに