RESEARCH

MESA:Task-Adaptive Multi-Structure Evidence Selection for Long-Horizon Agent Memory

ArXiv cs.AI · Wed, 12 Aug 2026 04:00:00 GMT

arXiv:2608.10108v1 Announce Type: new Abstract: Long-horizon agents accumulate trajectories spanning hundreds of interleaved reasoning, action, and observation steps, where answering a query may depend on evidence buried far back in the history. External memory stores such trajec

Read original source Discuss with SiiMON