Line of inquiry
Inquiring lines›How can multi-agent systems achiev…›What conditions allow multi-agent…›this line of inquiry
Should agents compress episodic memory or retain raw interaction histories?
A broader line of inquiry — a family of 50 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 50
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Why do agents ignore condensed experience in favor of raw data?
- Can agents improve if we constrain how much history they retain?
- Should agents update memory after every turn or batch process sessions?
- How should agents compress episodic interactions into working memory without accumulation?
- Do agents prefer raw experience over condensed summaries of past actions?
- What drives the choice between storing raw episodes versus abstracted rules?
- Does reducing interaction history cost agents performance on their tasks?
- Why did agents ignore condensed experience in the memory rewrites?
- Does compressing all past memories into one representation lose irretrievable details?
- 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?
- What details do high-level trajectory abstractions lose that state-grounded recall preserves?
- Do agents actually use condensed experience when it is their only available evidence?
- Can the same compress-then-act pattern work for agent state memory?
- How does memory folding enable agents to reconsider strategies mid-task?
- Why do continuously consolidated agent memories eventually degrade below no-memory baseline?
- Does selective history retrieval outperform full context inclusion in agent reasoning?
- Can agents compress long trajectories without losing critical decision context?
- Why does uniform memory consolidation sometimes degrade below the no-memory baseline?
- Can agents learn from their own experience without fine-tuning through episodic memory?
- What specific failure modes emerge when agents retrieve stale or contaminated memories?
- Why do successful and failed trajectories need different memory processing?
- How can agents distinguish over-generalized lessons from genuinely useful long-tail knowledge?
- Why does memory consolidation degrade agent performance below baseline?
- Why do agents systematically ignore condensed experience in their skill documents?
- How does indiscriminate memory injection cause multi-turn agent failures?
- Why do agents systematically underuse condensed experience in skill documents?
- How does memory extraction differ from retrieval in agent systems?
- Why does higher agent recall make forgetting problems harder?
- What shapes of memory help frozen agents improve without retraining?
- How should we evaluate agent memory if it folds into model computation instead of separate stages?
- Why does consolidating more state sometimes hurt performance below the no-memory baseline?
- When does persistent harmful memory create performance error floors?
- When should voice agents write new memories versus read existing ones?
- How should abstraction preserve applicability conditions when distilling experience?
- How do tool results and memory entries become injection vectors?
- What separates artifact recall from persistent memory commitment in agents?
- Does operator-conditioned memory let search compose learned behaviors more effectively?
- Why do agents ignore condensed experience even when it is the only evidence available?
- How do staleness, drift, and contamination each degrade agent memory differently?
- What makes naive memory consolidation regress below having no memory at all?
- How much actionable detail does condensation strip from raw experience?
- What happens when agents access interaction history beyond their assigned scope?
- How does bounded committed state prevent multi-turn agent failures better than transcript replay?
- Can memory consolidation fragility be detected and reversed during execution?
- How does raw conversation history differ from distilled memory profiles?
- How does structured environment state compare to transcript replay for multi-turn reasoning?
- How does continuous implicit memory formation differ from explicit memory encoding?
- What counts as scope when we restrict interaction history to agents?