Get Started
Home
Topics
Search
Library
Research questionHow can we tell whether a language diffusion model is memorizing or generalizing?A language diffusion model may recover stored training sequences or generate behavior that extends beyond them. The difficulty is determining which regime it occupies and measuring the transition between them.
AI
AI Memory
Diffusion Models
Evaluation & Benchmarks
Machine Learning
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Language Diffusion Models are Associative Memories Capable of Retrieving Unseen DataThe evidence concerns Uniform-based Discrete Diffusion Models and compares token recovery for training and unseen test examples. It reports conditional entropy of predicted token sequences as a probe, with the transition varying with training-dataset size.research paper · Sep 2, 2026
Related questions
How can we compare autoregressive and masked-diffusion language models without conflating formulation with architecture?Do language models generalize phonological rules or rely on memorized forms when selecting allomorphs?How can multilingual representation sharing be measured without confusing anisotropy with genuine cross-lingual structure?How can reinforcement learning post-training for diffusion models avoid objective mismatch and high-variance updates?