New agentic memory framework uses 118K tokens per query. LangMem burns through 3.26M.

NUS researchers' MRAgent framework reduces LLM agent memory retrieval to 118K tokens per query — vs. 3.26M for LangMem — ...
IEEE

Optimizing Query-by-Example Spoken Term Detection with Audio-to-Token Sequence Clustering and Query-Guided Retrieval

Abstract: Query-by-Example Spoken Term Detection (QbE-STD) retrieves relevant audio files corresponding to a spoken query, without relying on explicit word-level textual transcriptions. In ...