Research / Project detail

Memory, experience & agent systems

Advances Mem360 long-term memory and experience reuse, with research on agent workflows, memory safety, and experience serving.

MemoryAgentsEvaluation
Problem

How can agents preserve state, reuse experience from feedback, and account for the effect of runtime environments on performance?

Method

Mem360 concerns long-term user state and personalization; experience serving concerns task reflection and reusable feedback. Related research includes Harness-Bench, MemAudit, ArborMem, and experience serving.

Outcomes

The public Mem360 technical article reports LoCoMo Overall 76.46 and Temporal reasoning 83.32, among other results. Related research covers agent-workflow evaluation, memory safety, and experience serving.

Related publications

Related papers

Related coverage

Related coverage