Theme of inquiry
How do scale, context, and measurement affect capability and expertise?
A question within its area, explored through 3 lines of inquiry below — each a family of specific questions the research asks.
17 specific questions
- Why does memory consolidation degrade agent performance below baseline?
- Why does uniform memory consolidation sometimes degrade below the no-memory baseline?
- Why do continuously consolidated agent memories eventually degrade below no-memory baseline?
- What makes memory consolidation fragile compared to raw trajectory storage?
- Why does LLM memory consolidation regress below no-memory baselines?
- When does memory consolidation help agents instead of hurting performance?
- Why does consolidating more state sometimes hurt performance below the no-memory baseline?
27 specific questions
- How does durable memory quality shape agent performance over time?
- How should future memory systems control what gets written and trusted?
- Can agent-controlled memory management outperform fixed consolidation schedules?
- What makes timestamped knowledge repositories better than static memory?
- How do insert, forget, and merge operations maintain thought coherence over time?
- What drives the choice between storing raw episodes versus abstracted rules?
- What distinguishes formation, evolution, and retrieval as separate memory dynamics?
58 specific questions
- What persistent memory architectures best support storing precomputed inferences across sessions?
- Can precomputed inferences be stored in memory modules between model interactions?
- How do adaptive memory modules compare to feedback-based working memory for long context?
- Does including full context always degrade memory retrieval quality in practice?
- How do recurrent memory systems handle ultra-long context differently than attention?
- How do retention gates regularize forgetting across different sequence model architectures?
- How should memory systems split between short-term and long-term storage?