すべてのプロジェクト

jev-test

clduab11/jev-test

Pre-registered benchmark: can a 2B local model (Gemma 4 E2B) answer web questions without making things up when a decision model (TypeSafe Jev) makes every call? SearXNG for search, MemPalace for verbatim memory, seven arms including open local judges. Spec and thresholds fixed before any run.

Python
スター
1
フォーク
0

掲載理由

jev-testは検索・根拠選択・文の裏付けを別モデルのTypeSafe Jevで判定する研究用ハーネスで、READMEは判定付き構成が基準を下回ったとも記す。

関連プロジェクト

  • no-hallucination

    Three measured experiments on RAG hallucination: quote-checking, TypeSafe's Jev, and IBM's STAIR. 850+ graded questions, raw responses included.

    共通トピック: benchmark, hallucination, rag

  • jev-rag-benchmark

    Free English RAG benchmark for TypeSafe Jev 1.13 — frozen candidate pools, calibration, paired bootstrap CIs, $0 runs. Compared with OpenJev and NVIDIA cross-encoders.

    共通トピック: benchmark, rag

  • jev-ecosystem-research

    Jev (TypeSafe AI System One) 生态调研:开源复现规模、JevBench v1.4.2.2 全量解读、可获取性与许可证陷阱、router 生态、Jev+安全科研切入点

    共通トピック: benchmark, rag