What question did this study set out to answer?

The aim is to develop a compressed memory notation for AI agents that improves memory across sessions.

February 27, 2026Open Access

.ava: A Compressed Symbolic Notation for Persistent AI Agent Memory

Key Points

The aim is to develop a compressed memory notation for AI agents that improves memory across sessions.
Introduced .ava as a compressed symbolic notation for memory.
Formalized the grammar for .ava.
Conducted fine-tuning experiments on the Qwen2.5-1.5B model.
Achieved human-readable and editable memory representation.
Achieved a 2.65x compression ratio in tokens compared to French natural language.
Achieved 78.5% token accuracy in fine-tuning experiments.

Abstract

LLM agents operating continuously face a fundamental constraint: the finite context window limits memory between sessions. We present .ava, a compressed symbolic notation achieving 2.65x compression ratio in tokens compared to French natural language. Unlike vector memories, .ava is human-readable, editable, and version-controllable. We formalize its grammar, report fine-tuning experiments on Qwen2.5-1.5B (6 GB VRAM, 78.5% token accuracy), and propose a cognitive architecture including the concept of curvature. Also available in French: .ava : Une notation symbolique compressée pour la mémoire persistante d'agents IA

Read Full Paperexternally

Perguntar à IA

Bookmark

View Full Paper