Personalized presentation generation requires more than conditioning on a current prompt or template: agents must preserve stable user preferences across tasks, retain newly introduced preferences and constraints during multi-turn revision, and carry out local edits reliably. We propose MemSlides, a hierarchical memory framework for personalized presentation agents that separates long-term memory from working memory and further divides long-term memory into user profile memory and tool memory. User profile memory stores intent-conditioned profiles for round-0 personalization, working memory carries active preferences and session constraints across revision rounds, and tool memory stores reusable execution experience for reliable localized editing. MemSlides pairs this memory design with scoped slide-local revision, so targeted updates act on the smallest affected region instead of repeatedly regenerating the full deck. In controlled experiments, user profile memory improves persona-alignment judgments on a multi-persona, multi-intent profile bank, tool-memory injection improves closed-loop modify behavior in diagnostic matched-pair settings, and qualitative cases illustrate working memory's ability to carryover preferences. Taken together, these results suggest that effective personalization in presentation authoring depends on separating persistent user profiles, session-level working memory, and reusable execution experience across generation and localized revision.
Alex: Welcome to another episode of ResearchPod. Today, we're looking at a research framework called MemSlides — a system designed to help AI agents create and edit slide presentations in a more consistent, personalised way.
Sam: So is this paper basically asking why current AI tools seem to have "amnesia" when we ask them to make a new slide deck?
Alex: That's a fair way to put it. Most AI tools treat every presentation as a blank slate. You might spend ten minutes telling it you prefer clean fonts, minimal bullet points, and a dark colour scheme — and then the next time you open a new deck, it's forgotten all of it. You're starting from scratch again.
Sam: That's genuinely frustrating. It's like hiring an assistant who takes detailed notes, then shreds them at the end of every meeting. What do the researchers propose instead?
Alex: They propose what they call a "hierarchical memory" — which just means the AI keeps different kinds of notes for different purposes. Think of it in three layers. The first is a long-term style guide: your general preferences that rarely change, like always using your company's brand colours. The second is a log of past fixes — things that went wrong before and how they were corrected. And the third is a short-term working note just for the current deck, capturing any specific constraints for this particular project.
Sam: So the memory is organised by how long each piece of information stays relevant?
Alex: Exactly. Long-term memory holds the habits that persist across every project. Working memory holds the details that only matter right now. The key insight is that mixing those two things up is what causes the AI to behave inconsistently — it either forgets your standing preferences or it drags old, project-specific details into a new context where they don't belong.
Sam: And how does this change the way the AI actually makes edits? If I ask it to fix one slide, does it still rewrite the whole deck?
Alex: That's one of the central problems the paper addresses. Most systems, when asked to change something, will regenerate large portions of the presentation — which sounds helpful but often introduces new errors or quietly overrides your established style. The researchers instead use what they call "scoped slide-local revision." The idea is to identify the smallest region of the deck that actually needs to change, and only touch that.
Sam: So it's like using a fine brush to fix one corner of a painting, rather than repainting the whole canvas because one detail was off.
Alex: That's a good way to think about it. And the reason this matters is something the paper calls drift. Every time an AI rewrites more than it needs to, there's a chance it subtly shifts away from your original style. Keep the edits small and targeted, and the deck stays coherent over time.
Sam: How does the system actually decide what counts as "the smallest region"? Is there a structured process for that?
Alex: There is. They use what they call a "Plan-Act-Guard" pipeline — three steps that happen in sequence for every edit. First, the system draws up something like a contract: a precise description of what needs to change and what must not be touched. Then it carries out the edit. Then — and this is the part that makes it reliable — a separate checking step verifies that the change was applied correctly and that nothing else was accidentally altered.
Sam: So the "Plan" sets the boundaries, the "Act" makes the change, and the "Guard" is the quality check that catches anything that slipped through.
Alex: Precisely. And the reason that three-step structure matters is that it separates intention from execution. The system commits to a plan before it acts, which makes it much harder for the edit to quietly expand beyond its original scope. It's a bit like a surgeon marking exactly where to cut before picking up the scalpel — the discipline happens before the action, not during it.
Sam: It sounds like the whole framework is less about raw capability and more about discipline — giving the AI a clear sense of what to pay attention to and what to leave alone.
Alex: That's the central argument of the paper, yes. The researchers suggest that the reliability problem with current AI presentation tools isn't primarily about the AI being insufficiently powerful. It's about the AI lacking the organisational structure to manage context well. A persistent memory of your preferences, combined with a disciplined approach to making only the changes that are needed — that combination is what makes the tool feel genuinely useful over time, rather than impressive once and frustrating thereafter.
Sam: So it reframes slide editing as an ongoing, iterative conversation rather than a one-shot task you have to restart every time.
Alex: That's a good summary. And it points to a broader shift in how we might think about AI assistants — not as tools that generate things on demand, but as systems that build up a working understanding of you and your work over time. Thanks for listening to ResearchPod.