今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | SkillShapley: Boundary-Adaptive Shapley Valuation for Skill Step Attribution in LLM Agents | arXiv | 2026-08-13 | AI-Agent | 高 |
| 2 | Validation-Centric AI-Assisted GPU Porting of a 250,000+ Line Legacy Weather Simulation Code | arXiv | 2026-08-13 | AI-Agent | 高 |
重点论文与技术动态
1. SkillShapley: Boundary-Adaptive Shapley Valuation for Skill Step Attribution in LLM Agents
- 来源:arXiv
- 日期:2026-08-13
- 作者/机构:Chang Liu, Yuqi Zhang, Yiman Zhong, Boyi Liu, Hengjun Wang, Shuyue Wei
- 主题标签:
AI-Agent,arXiv - 推荐等级:高
- 分类:cs.AI
一句话结论
SkillShapley框架通过边界自适应Shapley估值,有效量化了LLM代理技能中各步骤的贡献,为技能创建提供关键指导。
核心内容
- 现有agent skills主要依赖人工或执行痕迹创建,缺乏对步骤贡献的量化理解,存在开放问题。
- SkillShapley采用两阶段框架:先识别信息性联盟区域,再自适应采样新联盟,利用离散化奖励导致性能悬崖和步骤交互加性特性。
- SkillsBench实验验证,该框架能高效识别高/低价值技能步骤,提供技能优化关键见解。
方法与数据
- 方法:基于Shapley值的两阶段框架,基于实证洞察(离散化奖励和加性交互)设计。
- 数据:在SkillsBench数据集上进行实验。
价值判断
- 值得关注:解决技能步骤贡献量化问题,对提升LLM代理执行复杂任务(如编码)效率至关重要。
- 可复用点:框架可迁移至其他代理技能创建,提供步骤级归因方法。
- 局限/待核查:摘要未明确提及局限。
摘要
Agent skills are crucial external instructions that enable language agents to execute long procedural tasks such as coding or document processing. Existing agent skills are primarily created through human manual crafting or agent execution traces, with limited understanding of how each step contributes to overall skill performance on specific tasks; i.e., there remains an open problem in quantifying the contribution of individual steps within an agent skill. To address this issue, we first model skill-step attribution as a Shapley value-based contribution estimation problem, and then propose SkillShapley, a step-level attribution framework for agent skills. Notably, SkillShapley operates in two phases, motivated by key empirical insights, i.e., discretized benchmark rewards that create sharp performance cliffs, and step interactions that are largely additive rather than synergistic. Specifically, it first identifies informative coalitional regions, and then adaptively samples new coalitions that can yield reusable marginal evidence. Experiments on skills from the widely adopted SkillsBench demonstrate that our SkillShapley can effectively and efficiently identify high- or low-value skill steps, providing several key takeaways for agent skill creation.2. Validation-Centric AI-Assisted GPU Porting of a 250,000+ Line Legacy Weather Simulation Code
- 来源:arXiv
- 日期:2026-08-13
- 作者/机构:Tetsuya Hoshino, Masaya Kato, Kazuhisa Tsuboki, Daichi Mukunoki, Takahiro Katagiri, Toshihiro Hanawa
- 主题标签:
AI-Agent,arXiv - 推荐等级:高
- 分类:cs.DC
一句话结论
AI辅助GPU端口化需以验证为中心设计工作流,确保科学有效性不被破坏并实现高效加速。
核心内容
- 工作流利用AI代理提取OpenMP区域、生成dump基准、应用OpenACC转换,并通过元素级比较和应用级验证确保结果一致。
- 在CReSS案例中,为162个目标内核生成数值验证GPU实现,应用级加速5.1倍,开发成本合理。
- 检测到5个内核的浮点和内在函数差异导致数值偏差(如阈值敏感分支发散),提供开发者反馈。
方法与数据
- AI代理驱动的验证工作流,管理会话上下文与状态重建。
- 摘要未明确具体数据集细节。
价值判断
- 值得关注:科学应用端口化必须优先验证科学有效性,避免数值偏差导致可信度损失。
- 可复用点:dump-based验证机制和差异检测流程可推广至大型遗留科学代码。
- 局限/待核查:小静态分析遗漏的恢复成本高,需优化静态分析精度以降低开发复杂性。