ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?

• 类型：arxiv
• 标识：2608.03874
• 链接：https://arxiv.org/abs/2608.03874
• 主分类：agent
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：ContinualSkillBench is introduced, a dynamic evaluation framework for in-context continual skill learning that shows that current in-context skill evolution mechanisms can support continual adaptation, but still struggle to consistently consolidate experience into robust and transferable skills.
• 待LLM分类：否
• 标题中文：ContinualSkillBench：LLM Agent 能否真正进化其能力？
• TLDR中文：提出 ContinualSkillBench，一个面向 in-context 持续 skill 学习的动态评估框架，表明当前 in-context skill 进化机制能够支持持续适应，但仍难以稳定地将经验整合为鲁棒且可迁移的 skill。
• 来源文件：
• /inbox/tom/_candidates/2026-08-05-agent-rag-longcontext-candidates.json
• [S2 enrich]