EPS
← All batches·2605.18813

Composition of Memory Experts for Diffusion World Models

topic: current_projecttop score: 100released: 2026-05-20first surfaced: 2026-05-20arXivPDFthreats2026-05-20

Authors: Sebastian Stapf, Pablo Acuaviva Huertos, Aram Davtyan et al.

arXiv · PDF

Summary

arXiv:2605. 18813v1 Announce Type: new Abstract: World models aim to predict plausible futures consistent with past observations, a capability central to planning and decision-making in reinforcement learning.

Relevance

Read next because Composition of Memory Experts for Diffusion World Models overlaps with clean result "LoRA persona trained on alone emits at 23.5% when a co-trained partner learns ..., vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rate, without, position, contexts, capability, test, model. Source: arxiv cs.LG (Machine Learning).

Threat model

Potential threat/caveat for clean result "LoRA persona trained on alone emits at 23.5% when a co-trained partner learns ..., vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.

Abstract

arXiv:2605.18813v1 Announce Type: new Abstract: World models aim to predict plausible futures consistent with past observations, a capability central to planning and decision-making in reinforcement learning. Yet, existing architectures face a fundamental memory trade-off: transformers preserve local detail but are bottlenecked by quadratic attention, while recurrent and state-space models scale more efficiently but compress history at the cost of fidelity. To overcome this trade-off, we suggest decoupling future-past consistency from any single architecture and instead leveraging a set of specialized experts. We introduce a diffusion-based framework that integrates heterogeneous memory models through a contrastive product-of-experts formulation. Our approach instantiates three complementary roles: a short-term memory expert that captures fine local dynamics, a long-term memory expert that stores episodic history in external diffusion weights via lightweight test-time finetuning, and a spatial long-term memory expert that enforces geometric and spatial coherence. This compositional design avoids mode collapse and scales to long contexts without incurring a quadratic cost. Across simulated and real-world benchmarks, our method improves temporal consistency, recall of past observations, and navigation performance, establishing a novel paradigm for building and operating memory-augmented diffusion world models.