遇见数据集

The Recall Trap: reproducibility artifact (code, served packs, predictions, grading reports, and pre-registration records)

收藏
Zenodo2026-08-10 更新2026-08-13 收录
官方服务:

资源简介:

Reproducibility artifact for "The Recall Trap: A Recall-Maximizing Retriever Configuration Reduces Issue Resolution in Fixed-Budget Code Context." Contains the single-shot and grading harnesses, served 12-slot context packs, per-instance model predictions and official SWE-bench Verified / SWE-PolyBench grading reports for every arm, the raw-response logs for the pre-registered gpt-5.6-sol logged replication, and all pre-registration records with a provenance ledger (PREREG-LEDGER.md) mapping each pre-registered claim to a git commit. Regenerable per-instance checkouts and Docker images are excluded by design; results rebuild from the served packs. SHA-256 of the archive is recorded in MINT-MANIFEST.md.

提供机构:
Zenodo
创建时间:
2026-08-10
二维码
社区交流群
二维码
科研交流群
商业服务