Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation
- 类型:arxiv
- 标识:2607.05382
- 链接:https://arxiv.org/abs/2607.05382
- 主分类:agent
- 形态:method
- 被引:3
- 被引来源:Semantic Scholar
- S2被引:3
- OpenAlex被引:0
- 影响力被引:0
- TLDR:This work traces the root cause of naive search to a generator-specific, evolving knowledge boundary: the divide between what a generator can internalize through training and what must remain in external context, and shows that it is discoverable through a teach-then-search co-training framework.
- OpenAlex ID:W7167620962
- OpenAlex DOI:10.48550/arxiv.2607.05382
- DOI:10.48550/arxiv.2607.05382
- DOI来源:OpenAlex
- 开放获取:green
- 开放获取链接:https://doi.org/10.48550/arxiv.2607.05382
- OpenAlex更新:2026-07-19
- 待LLM分类:否
- 标题中文:Search Beyond What Can Be Taught:Agentic 视觉生成中的知识边界演化
- TLDR中文:本研究将朴素搜索的根因追溯到生成器特有的、可演化的知识边界——即生成器经训练可内化的内容与必须保留于外部上下文的内容之间的鸿沟,并表明该边界可通过"先教后搜"协同训练框架被有效发现。
- 副分类:engineering
- 来源文件:
- /inbox/tom/_candidates/2026-07-16-agent-rag-longcontext-candidates.json
- [S2 enrich]
- [OpenAlex backfill]