Chinese contentFRESH：用异构图记忆让小模型工具调用智能体从失败中学习
小模型跑工具智能体老犯重复调用、提前写入这类错，这篇把失败经验做成异构图记忆来避免，τ-Bench 和 AppWorld 都验证有效，做本地部署的可以看看。
#τ-Bench
2 stories · 0 in 7 days · 0 in 30 days
Stories, products and papers tagged "τ-Bench", newest first.
Sep 24
Sep 1
Hidden-State Trajectories for Adaptive Reasoning
This method improves LLM reasoning efficiency across tasks, reducing costs while increasing success rates, making it valuable for developers working with constrained resources.