Row-level results for a controlled SkillsBench study of skill availability and presentation granularity (GPT-5.5 and DeepSeek V4-Flash)

Name: Row-level results for a controlled SkillsBench study of skill availability and presentation granularity (GPT-5.5 and DeepSeek V4-Flash)
Creator: Zenodo
Published: 2026-06-02T15:05:53+08:00

Zenodo2026-06-02 更新2026-06-05 收录

大型语言模型评估

技能基准测试

数据链接：

https://zenodo.org/doi/10.5281/zenodo.20512027 数据链接链接失效反馈

官方服务：

资源简介：

This dataset contains the row-level experimental results from a controlled study of skill availability and presentation granularity for large-language-model agents on a pinned SkillsBench subset. Contents:gpt55_results.jsonl: 900 row-level results for the GPT-5.5 configurationdeepseek_results.jsonl: 900 row-level results for the DeepSeek V4-Flash configuration

提供机构：

Zenodo

创建时间：

2026-06-02