Building a long-term memory benchmark for agents: what to add?

True_Mongoose_7073 · reddit · 2026-09-01

Developer is building a benchmark for agent long-term memory, currently featuring months of chat history, multilingual data, images, answer changes, and token usage tracking. Seeking feedback on missing features, especially edge cases encountered in real-world memory usage.

Original post →

More from coding & agent

coding & agent channel →