FRAME: Full Data Results
收藏官方服务：
资源简介：
Per-question records, model responses, LLM-judge verdicts, a checksum manifest, and verification scripts for FRAME, an evaluation control protocol for evidence rendering in memory/RAG evaluation (~259,000 calls, nine commercial LLMs). Version 2 adds the same-content rendering ablation and the oracle raw-excerpt baseline; see README.md for layout, scoring, and upstream licenses.
提供机构：
Zenodo创建时间：
2026-08-02
