EMNLP paper: language diffusion models act as associative memories
LucaAmb · x · 2026-11-03
A Bao Pham et al. EMNLP 2025 main-conference paper asks whether language diffusion models can recall sentences they've never seen. The finding: they behave as associative memories — as training data grows, unseen sentences become stable and recoverable from noise, and conditional entropy tracks the shift from memorization to generalization.
Paper and project page are public; collaborators include Matteo Negri, Dima Krotov, and Luca Ambrosini, who frames it as another chapter in the 'generative diffusion as associative memory' line of work.
More from Research
- GPT Astra-assisted proof improves Erdős Problem #357 bound, Lean-verified but not refereed — kevrussell · 2026-11-03
- RVCBench from UBC & Vector Institute stress-tests 18 voice cloning models across 18 robustness evaluations — jiqizhixin · 2026-11-03
- Counting mRNAs in living bacteria: the 2005 Golding paper that overturned the Poisson model of transcription — NikoMcCarty · 2026-11-03
- DeepSeek V4.1 Flash 1017 doubles ARC-AGI-3 score to 5.81%, leads open models — teortaxesTex · 2026-11-03
- Critics tear apart "Weight-Space Programming": Atari test spoon-feeds answers, Qwen case fails on Gemma — anshulkundaje · 2026-11-03
- Cohere paper: adjacent tokens share 20-45% of experts, making speculative decoding cheap even at batch size 1 — rohanpaul_ai · 2026-11-03