Human-Inspired Memory Architecture for LLM Agents
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910203496103936 |
|---|---|
| author | Kerestecioglu, Doga Robsky, Alexei Vasters, Clemens Sharma, Anshul Kesselman, Yitzhak |
| author_facet | Kerestecioglu, Doga Robsky, Alexei Vasters, Clemens Sharma, Anshul Kesselman, Yitzhak |
| contents | Current LLM agents lack principled mechanisms for managing persistent memory across long interaction horizons. We present a biologically-grounded memory architecture comprising six cognitive mechanisms: (1) sleep-phase consolidation, (2) interference-based forgetting, (3) engram maturation, (4) reconsolidation upon retrieval, (5) entity knowledge graphs, and (6) hybrid multi-cue retrieval. Each mechanism addresses a specific failure mode of naive memory accumulation. We introduce a synthetic calibration methodology that derives all pipeline thresholds without benchmark data exposure, eliminating a common source of evaluation leakage. We evaluate on two benchmarks. First, a VSCode issue-tracking dataset (13K issues, 120K events) where deduplication-based consolidation achieves 97.2% retention precision with 58% store reduction (+21.8 pp over baseline). Second, the LongMemEval personal-chat benchmark where we conduct the first streaming M-tier evaluation (475 sessions, ~540K unique turns). At a 200K-token context budget, our pipeline matches raw retrieval accuracy (70.1% vs. 71.2%, overlapping 95% CI) while exposing a tunable accuracy/store-size operating curve. At S-tier scale (50 sessions), dedup-based consolidation yields a +13.3 pp improvement in preference recall. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_08538 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Human-Inspired Memory Architecture for LLM Agents Kerestecioglu, Doga Robsky, Alexei Vasters, Clemens Sharma, Anshul Kesselman, Yitzhak Artificial Intelligence Computation and Language Information Retrieval Machine Learning Current LLM agents lack principled mechanisms for managing persistent memory across long interaction horizons. We present a biologically-grounded memory architecture comprising six cognitive mechanisms: (1) sleep-phase consolidation, (2) interference-based forgetting, (3) engram maturation, (4) reconsolidation upon retrieval, (5) entity knowledge graphs, and (6) hybrid multi-cue retrieval. Each mechanism addresses a specific failure mode of naive memory accumulation. We introduce a synthetic calibration methodology that derives all pipeline thresholds without benchmark data exposure, eliminating a common source of evaluation leakage. We evaluate on two benchmarks. First, a VSCode issue-tracking dataset (13K issues, 120K events) where deduplication-based consolidation achieves 97.2% retention precision with 58% store reduction (+21.8 pp over baseline). Second, the LongMemEval personal-chat benchmark where we conduct the first streaming M-tier evaluation (475 sessions, ~540K unique turns). At a 200K-token context budget, our pipeline matches raw retrieval accuracy (70.1% vs. 71.2%, overlapping 95% CI) while exposing a tunable accuracy/store-size operating curve. At S-tier scale (50 sessions), dedup-based consolidation yields a +13.3 pp improvement in preference recall. |
| title | Human-Inspired Memory Architecture for LLM Agents |
| topic | Artificial Intelligence Computation and Language Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2605.08538 |