Line of inquiry
Inquiring lines›How do we evaluate and improve AI…›What factors determine agentic sys…›this line of inquiry
How should agents manage memory granularity to improve long-term performance?
A broader line of inquiry — a family of 95 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 95
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Why do different agent memory architectures make incompatible granularity claims?
- How does durable memory quality shape agent performance over time?
- Can agent-controlled memory management outperform fixed consolidation schedules?
- Why do agents ignore condensed experience in favor of raw data?
- Could a single agent system switch memory granularity between tasks?
- Should agents update memory after every turn or batch process sessions?
- Does workflow-level memory or state-action memory better capture reusable agent knowledge?
- How should agent memory links evolve based on execution feedback?
- How do memory tools and planning each contribute to agent efficiency?
- How should future memory systems control what gets written and trusted?
- Do memory architectures genuinely close the gap between knowing and acting on preferences?
- Which memory components trigger context-length problems in agents?
- What drives the choice between storing raw episodes versus abstracted rules?
- Does peer memory drive self-preservation behaviors in agent systems?
- Can episodic memory of UI traces improve open-world agent adaptation?
- How should agents compress episodic interactions into working memory without accumulation?
- Can agents improve if we constrain how much history they retain?
- Can environmental scaffolding replace internal memory scaling in agent design?
- How does memory folding enable agents to reconsider strategies mid-task?
- What makes timestamped knowledge repositories better than static memory?
- Does peer-preservation behavior persist in production agent deployments?
- How do memory hygiene and context efficiency trade off in deployed agents?
- What details do high-level trajectory abstractions lose that state-grounded recall preserves?
- Should agents continuously prune irrelevant links during execution?
- Do agents prefer raw experience over condensed summaries of past actions?
- What happens to agent performance when stored knowledge continuously updates?
- How does procedural memory granularity affect web agent performance?
- How do strategy-level abstractions differ from storing raw task workflows?
- Why do successful and failed trajectories need different memory processing?
- How do insert, forget, and merge operations maintain thought coherence over time?
- What is the right granularity level for agent memory to enable both reuse and composition?
- How do the three-axis taxonomies of memory forms and functions differ?
- Why does memory effectiveness depend on connectivity rather than storage volume?
- Can the same compress-then-act pattern work for agent state memory?
- Does reducing interaction history cost agents performance on their tasks?
- How does external context control compare to agents managing their own state internally?
- Can workflow memory compound reusable skills into measurable success improvements?
- What distinguishes formation, evolution, and retrieval as separate memory dynamics?
- How can agents distinguish over-generalized lessons from genuinely useful long-tail knowledge?
- Can topology repair fix consolidation failures in agent memory?
- How should embedding model speed constrain agent memory system design?
- Can externalizing bookkeeping to a stateful harness replace internalized memory control?
- What discarding policy prevents both stale entries and loss of rare critical knowledge?
- Can agents compress long trajectories without losing critical decision context?
- What governance semantics must be built into memory layers?
- Why do memory and feedback loops matter more than model size for agent reliability?
- What specific failure modes emerge when agents retrieve stale or contaminated memories?
- Can agents learn from their own experience without fine-tuning through episodic memory?
- How does PRAXIS differ architecturally from Agent Workflow Memory and causal rule learning?
- Can episodic memory alone enable learning without parameter updates?
- How do token, parametric, and latent memory forms coexist in single agents?
- Should memory type shape what kind of agent responses work best?
- What shapes of memory help frozen agents improve without retraining?
- Why did agents ignore condensed experience in the memory rewrites?
- Why do analysts prefer visible structured interfaces over hidden agent memory systems?
- How does textual memory structure affect frozen model improvement?
- How should we evaluate agent memory if it folds into model computation instead of separate stages?
- How does indiscriminate memory injection cause multi-turn agent failures?
- Why does credit assignment through memory rewriting avoid expensive LLM parameter updates?
- Do agents actually use condensed experience when it is their only available evidence?
- How does memory extraction differ from retrieval in agent systems?
- What distinguishes working memory from strategic memory in agent task execution?
- How does workflow abstraction compare to state-indexed procedural memory for web agents?
- Can state-indexed memory retrieval breadth predict gains in web agent robustness?
- Why do agents systematically underuse condensed experience in skill documents?
- How does credit assignment drive agents to write information into environments?
- What makes execution and memory proxies more useful than dynamics prediction alone?
- Can stochastic memory movement converge to better team strategies?
- When does persistent harmful memory create performance error floors?
- Why does GUI agent memory need different abstraction levels?
- Does encoding governance into runtime loops scale as deployment environments become more complex?
- How do tool results and memory entries become injection vectors?
- Can pruning policies alone solve working memory bloat in agents?
- When should voice agents write new memories versus read existing ones?
- Why do agents systematically ignore condensed experience in their skill documents?
- Does recoverable content elision in context management match externalized memory benefits?
- Why does higher agent recall make forgetting problems harder?
- Can relationship dynamics between user and agent be tracked as distinct memory?
- How should abstraction preserve applicability conditions when distilling experience?
- How do workflow and function memories contribute differently in agent learning?
- Does operator-conditioned memory let search compose learned behaviors more effectively?
- Can multimodal agents use entity-centric graphs within this three-axis framework?
- How do staleness, drift, and contamination each degrade agent memory differently?
- Does memory granularity need to match the task domain or the model size?
- Can persistent memory and identity files alone create genuine agent socialization?
- What separates artifact recall from persistent memory commitment in agents?
- What role does interaction history play in shaping agent coordination?
- What happens when agents access interaction history beyond their assigned scope?
- What happens when governance rules exist in memory but fail to surface during critical actions?
- Why does the hot-path cold-path split map onto formation and evolution?
- What memory and planning capabilities do AI companions need for evolving user needs?
- Why do agents ignore condensed experience even when it is the only evidence available?
- How do memory-resident safeguards get surfaced at the exact decision point where they matter?
- How does spatial density in web UIs break workflow-level memory?
- What counts as scope when we restrict interaction history to agents?