Memanto: Typed Semantic Memory with Information-Theoretic Retrieval for Long-Horizon Agents
Seyed Moein Abtahi, Rasa Rahnema, Hetkumar Patel, Neel Patel, Majid Fekri, Tara Khani · 2026
Read Memanto: Typed Semantic Memory with Information-Theoretic Retrieval for Long-Horizon Agents by Seyed Moein Abtahi, Rasa Rahnema, Hetkumar Patel, Neel Patel, Majid Fekri, Tara Khani (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-22085.txt.
Description
Abtahi and five colleagues present Memanto, a memory layer for agent AI that drops the idea that graph complexity is needed for true-to-life agent memory. Memanto combines a typed meaning store with thirteen set groups, auto clash fixing, and time versions. On LongMemEval and LoCoMo, Memanto achieves top accuracy of 89.8 percent and about 87 percent. It beats all tested hybrid graph and vector systems with one search, no intake cost, and much lower running complexity.
Catalog record
| CREATOR | Seyed Moein Abtahi, Rasa Rahnema, Hetkumar Patel, Neel Patel, Majid Fekri, Tara Khani |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55165 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.22085