newsThe Register AITrust 72 · OutletPublished yesterdayLive · 21h ago
AI models get convenient amnesia about source material as they grow, MIT boffins find
Attributing diffusion model output to a specific input becomes more difficult with more training data
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 56%From Global to Factor-Wise Expert Composition in Discrete Diffusion Models →
- PossiblePossibly related (embedding) · 53%An Empirical Study of Training Pixel-Space Text-to-Image Diffusion Models →
- PossiblePossibly related (embedding) · 53%Steering Optimisation Trajectories in Diffusion Representation Learning →
- PossiblePossibly related (embedding) · 52%Adaptive Block Diffusion: Resolving Training-Inference Mismatch in Diffusion Language Models →
- PossiblePossibly related (embedding) · 51%Trace-Based On-Policy Distillation for Masked Diffusion Language Models →
Covers
paperFrom Global to Factor-Wise Expert Composition in Discrete Diffusion ModelspaperAn Empirical Study of Training Pixel-Space Text-to-Image Diffusion ModelspaperSteering Optimisation Trajectories in Diffusion Representation LearningpaperAdaptive Block Diffusion: Resolving Training-Inference Mismatch in Diffusion Language ModelspaperTrace-Based On-Policy Distillation for Masked Diffusion Language Models
Related across the graph
paperSteering Optimisation Trajectories in Diffusion Representation LearningpaperAn Empirical Study of Training Pixel-Space Text-to-Image Diffusion ModelspaperFrom Global to Factor-Wise Expert Composition in Discrete Diffusion ModelspaperAdaptive Block Diffusion: Resolving Training-Inference Mismatch in Diffusion Language ModelspaperTrace-Based On-Policy Distillation for Masked Diffusion Language Models