LLM智能体自我进化可能反噬，预承诺门控机制防止技能污染，揭秘新解法。
arXiv:2608.05810v1 Announce Type: new Abstract: Self-evolving agents accumulate capability by distilling reusable skills from their execution trajecto…
多智能体与人类协同，门控架构助力经济理论创新——让AI成为经济学家的得力助手。
arXiv:2607.21268v1 Announce Type: cross Abstract: In many social-science research tasks, such as economics, LLM-based agents must produce outputs for …
开源可解释门控机制，专治LLM提示注入攻击，安全防护新思路
Article URL: https://github.com/cgrtml/reasongate Comments URL: https://news.ycombinator.com/item?id=48941051 Points: 5 # Comments: 0
推翻重排序必有益的假设，提出无训练门控机制：只有不确定性高时再重排序，反而提升性能。
arXiv:2606.31087v1 Announce Type: cross Abstract: Few-shot selection typically assumes that reranking retrieved examples always improves performance. …
一种从稠密到稀疏的持续训练新方法，利用预测器门控实现高效稀疏化，针对大语言模型场景。
arXiv:2606.10722v1 Announce Type: new Abstract: We study dense-to-sparse continual training as a way to construct channel-sparse large language models…
提出动态logit级门控机制，优化LLM专家路由效率，大幅降低推理成本。
arXiv:2606.04378v1 Announce Type: new Abstract: Leveraging multiple specialized LLMs can combine complementary strengths, but existing approaches trad…
用特权信息生成密集token级监督，提升LLM推理能力的新蒸馏方法。
arXiv:2605.28791v1 Announce Type: cross Abstract: On-policy self-distillation (SD) improves LLM reasoning by using teacher-side privileged information…
专为AI代理设计的门控系统GateGraph，能在动作执行前决定是否放行，保障安全与可控性。
Article URL: https://github.com/humancoreai/Gategraph Comments URL: https://news.ycombinator.com/item?id=48251003 Points: 1 # Comments: 0