Salesforce 提出 HASP 框架：用技能程序拦截智能体失败动作
Salesforce 这个 HASP 有点意思：把失败教训写成程序，在智能体要出错时自动纠偏，准确率直接从 31.2% 拉到 51%，还省 token。
Salesforce Research 发布 HASP 框架，把智能体过往的失败经验转成可执行的 Program Functions，在容易出错的环节修改下一步动作或补充纠正上下文。在网页搜索推理任务上，HASP 将准确率从 31.2% 提升到 51.0%，同时比 ReAct 智能体少用约 15% token，延迟和成本也更低。技能库会随新失败不断扩充，每个候选程序都需通过代码验证和教师模型审查；去掉这一过滤后，准确率会从 60.3% 跌到 36.3%。每次干预都会记录触发条件、干预前后的动作和原因，便于追溯。
Harnessing LLM Agents with Skill Programs (HASP): https://t.co/vodO1l0eQf
HASP turns lessons from past agent failures into executable Program Functions that step in at failure-prone moments to revise the next action or add corrective context.
→ About 15% fewer tokens, lower latency, and lower cost per query than a ReAct agent, with accuracy rising from 31.2% to 51.0% on web-search reasoning → Every intervention logs its trigger, the action before and after, and the reason → The skill library grows from new failures, and each candidate must pass code validation and teacher review → Without that filtering, accuracy falls from 60.3% to 36.3%
#FutureOfAI #EnterpriseAI #AgenticAI
