Agentsensus reduces memory duplication in multi-character story simulations
Agentsensus: Consensus-Compressed Shared Memory for Multi-Agent Story Worlds
Artificial IntelligenceComputation and LanguageMultiagent Systems
Summary
Simulating stories with many characters usually means each character remembers the same event separately, causing lots of data copies. The authors created Agentsensus, which stores shared memories of events once for all characters who saw them, linking related memories together. This method saves storage space by 22-44% compared to typical designs, while keeping or improving story quality. The key trick is merging memories so multiple characters share records, which grows as the story runs longer.
What this means in practice
- •For game developers: Build story-driven games that handle many characters with less memory by sharing common event records among characters.
- •For simulation engineers: Create multi-agent simulations of social or historical events with efficient memory use due to unified and linked event storage.
Authors
Yu Pan
Abstract
A agentic story world is a dynamic system simulating who learned what, when, and from whom -- yet the standard design gives each character a private memory stream. A shared event is therefore stored once per witness, large duplication will be incurred in terms of storage. We present Agentsensus, a story-world simulation framework in which there is an unified long-term memory. Records of the same event merge into one owned by all its witnesses, and semantically relevant memory records are linked. We evaluate on four worlds -- two classical Chinese novels, Hamlet, and a real-world conflict timeline -- run for 40 to 80 rounds against three per-character memory designs under an equal-granularity protocol. Agentsensus writes 22-44% fewer entries than the closest baseline and is the only design whose memory becomes shared (14-28% of records held by more than one character, some by 10) and linked (94-99%), at judged simulation quality indistinguishable or even better than the baselines. An ablation attributes this to the merge itself: disabling it multiplies the store by 3.1x and takes sharing to exactly zero. Sharing also compounds with the horizon rather than saturating early, rising 6% to 9% to 14% as one world is re-run at 10, 20 and 40 rounds.