Tag: Adaptive Systems
All the articles with the tag "Adaptive Systems".
-
Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreaks
ASTRA introduces an efficient defense for Vision Language Models by adaptively steering activations away from adversarial directions using image attribution, achieving state-of-the-art performance in mitigating jailbreak attacks with minimal impact on benign utility and high inference efficiency.
-
AdaptMI: Adaptive Skill-based In-context Math Instruction for Small Language Models
本文提出AdaptMI和AdaptMI+自适应方法,通过基于奖励模型检测问题难度并针对困难问题选择技能-based in-context示例,提高小语言模型在数学推理任务中的性能,同时避免认知过载。
-
Plan-and-Act: Improving Planning of Agents for Long-Horizon Tasks
本文提出PLAN-AND-ACT框架,通过分离规划和执行模块、利用合成数据训练和动态重规划,提高LLM代理在复杂长期任务中的性能,并在web导航基准上达到state-of-the-art结果。
-
Adaptive Layer-skipping in Pre-trained LLMs
本文提出FlexiDepth方法,通过插件式路由器和适配器实现预训练LLM的自适应层跳过,提高计算效率同时保持生成性能,并通过实验揭示了token类型对计算需求的影响。
-
Streaming, Fast and Slow: Cognitive Load-Aware Streaming for Efficient LLM Serving
本文提出基于认知负载的适应性流式传输框架,用于优化 LLM 服务,通过动态调整输出速度减少计算资源消耗高达 16.8%,同时维持用户满意度。