Search-on-Graph-R1: Training Large Language Models to Search Knowledge Graphs with Reinforcement Learning
Jia Ao Sun, Hao Yu, Fengran Mo, Zhan Su, Yuchen Hui
実装難易度
Hard
推論・学習コスト
High
想定用途
QA
概要
Knowledge graph question answering (KGQA) requires navigating from topic entities to an answer several relations away. Recent methods prompt a frontier LLM to explore the graph through a retrieval tool, but their reliance on frontier-scale inference makes them costly to deploy. We present Search-on-Graph-R1 (\sogrone{}), which internalizes this navigation into a compact 8B model through supervised…
何が新しいか
Knowledge graph question answering (KGQA) requires navigating from topic entities to an answer several relations away. Recent methods prompt a frontier LLM to explore the graph through a retrieval…
何に使えるか
QA
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
Debate-on-Graph: Reliable and Adaptive Reasoning of Large Language Model on Uncertain Knowledge Graph
Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing. However
RIMS: Preference Optimization via Smoothed Multi-pair Aggregation for Small-Scale LLM Retrieval-Augmented Generation
Small-scale language models (SLMs) are attractive for retrieval-augmented generation (RAG) in resource-constra
VIABench: A Comprehensive Video Benchmark Collected from Blind Individuals for Visual Impairment Assistance
Visually impaired individuals (VIIs) encounter significant daily challenges due to limited access to visual in
Dual Adversarial Fine-tuning for Enhancing Robustness of Large Vision Language Model
この研究では、大規模視覚言語モデル(LaVaとGPT-4V)が不正データに弱い問題に対処し、視覚的情報の安全性を向上させることができる方法が提案されている。