Facts as First Class Objects: Knowledge Objects for Persistent LLM Memory

TL;DR

This paper compares in-context memory with Knowledge Objects (KOs) for persistent LLM memory, showing KOs are more accurate, cost-effective, and robust against capacity and compaction issues, with improved multi-hop reasoning.

Contribution

It introduces Knowledge Objects as a discrete, hash-addressed memory system for LLMs, demonstrating superior accuracy and efficiency over in-context memory, and provides a benchmark suite for evaluation.

Findings

KOs achieve 100% accuracy across all tested conditions.

KOs are 252 times more cost-effective than in-context memory.

Embedding retrieval fails on adversarial facts, highlighting limitations.

Abstract

Large language models increasingly serve as persistent knowledge workers, with in-context memory - facts stored in the prompt - as the default strategy. We benchmark in-context memory against Knowledge Objects (KOs), discrete hash-addressed tuples with O(1) retrieval. Within the context window, Claude Sonnet 4.5 achieves 100% exact-match accuracy from 10 to 7,000 facts (97.5% of its 200K window). However, production deployment reveals three failure modes: capacity limits (prompts overflow at 8,000 facts), compaction loss (summarization destroys 60% of facts), and goal drift (cascading compaction erodes 54% of project constraints while the model continues with full confidence). KOs achieve 100% accuracy across all conditions at 252x lower cost. On multi-hop reasoning, KOs reach 78.9% versus 31.6% for in-context. Cross-model replication across four frontier models confirms compaction loss is architectural, not model-specific. We additionally show that embedding retrieval fails on adversarial facts (20% precision at 1) and that neural memory (Titans) stores facts but fails to retrieve them on demand. We introduce density-adaptive retrieval as a switching mechanism and release the benchmark suite.