MemHarness: Memory Is Reconstructed, Not Replayed
実装難易度
Easy
推論・学習コスト
High
想定用途
異常検知
概要
Retrieving past experiences has become a common strategy to enhance large language model agents. However, most existing memory-augmented agents treat retrieved experiences as static records to be replayed verbatim, injecting them into the context regardless of whether they align with the agent's current situation. This ``replay'' paradigm ignores the gap between the abstract, general nature of…
何が新しいか
Retrieving past experiences has become a common strategy to enhance large language model agents. However, most existing memory-augmented agents treat retrieved experiences as static records to be…
何に使えるか
異常検知
実装情報
- Hugging Face 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
OKHugging Face
検証しやすさ
OK実装またはモデル配布ページから試せる可能性が高いです。
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
Easy - 実装またはモデル配布ページから試せる可能性が高いです。
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: 推論だけで試せ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
Intertemporal Preference Steering in Qwen3 via Contrastive Activation Addition
We study linear representations of temporal horizon in the large language model Qwen3-32B and use them to chan
Failure-Informed Image Self-Augmentation for Multimodal Large Language Model Self-Improvement
Multimodal large language models (MLLMs) have achieved remarkable performance across vision-language tasks, bu
ToolLIFT: Lifting Tool-Specific Trajectories into Function-Level Graphs for Generalizable Tool Planning
歴史的なtool用途のトラジケティを用いてFunction-Levelグラフを構築し、工具の指定からでも汎用的にコードを生成すること。
LDU-Bench: Multimodal LLM Evaluation for Lithography Defect Understanding under Layout-Varying Circuit Backgrounds
Multimodal large language models have demonstrated strong defect recognition capability in industrial anomaly