Search papers, labs, and topics across Lattice.
This paper introduces MemoryLACE (MemLACE), a lightweight memory framework designed to enhance long-term memory management in LLMs by explicitly modeling the lifecycle of textual evidence through sparse merge, supersession, and contradiction relations. By reconstructing relation-aware evidence units, MemLACE effectively distinguishes between current, historical, supporting, and conflicting information, leading to improved reasoning capabilities. The framework outperforms existing methods, achieving a 66.6% reduction in runtime on the BEAM benchmark while maintaining the highest performance in same-backbone comparisons, highlighting the importance of lifecycle awareness in memory systems.
Explicitly modeling the lifecycle of textual evidence can dramatically enhance long-term memory reasoning in LLMs while cutting runtime significantly.
Long-term LLM agents must preserve information across interactions while distinguishing repeated evidence, historical states, updates, and unresolved contradictions. Existing textual memory systems retrieve semantically relevant memories efficiently but often leave these relationships implicit, whereas richer structured approaches model them through global graphs, hierarchical abstractions, or reflection at greater complexity. We introduce MemoryLACE (MemLACE), a lightweight memory framework that explicitly models the lifecycle of textual evidence through sparse merge, supersession, and contradiction relations while preserving atomic natural-language memories and their provenance. Rather than retrieving memories independently, MemLACE reconstructs relation-aware evidence units that expose current, historical, supporting, and conflicting evidence for downstream reasoning. Across BEAM and StructMemEval, using open-weight and proprietary LLM backbones, MemLACE achieves the highest overall performance in same-backbone comparisons while reducing end-to-end runtime on BEAM by 66.6% relative to Hindsight, the strongest reported reflective-memory baseline. Ablation studies identify lifecycle expansion and temporal awareness as the principal contributors to these gains. Together, the results demonstrate that explicitly modeling the local lifecycle of textual evidence is sufficient to substantially improve long-term memory reasoning without requiring comprehensive knowledge graphs or global reflection.
