Clock Diffusion: Efficient Semi-Autoregressive Continuous Diffusion Language Models
cs.LG
Submitted: 2026-10-01
Updated: 2026-10-01
Code: https://github.com/Lightning-AI/lightning
Project page: http://skylion007.github.io/OpenWebTextCorpus
Sources
- Posterior Refinement: Fast Language Generation via Any-Order Flow Maps
- Set Diffusion: Interpolating Token Orderings Between Autoregression and Diffusion for Fast and Flexible Decoding
- Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models
- Accelerating Large Language Model Decoding with Speculative Sampling
- Analog Bits: Generating Discrete Data using Diffusion Models with Self-Conditioning
- LangFlow: Continuous Diffusion Rivals Discrete in Language Modeling
- Training Verifiers to Solve Math Word Problems
- Scaling Categorical Flow Maps
- Language Modeling with Hyperspherical Flows
- Continuous diffusion for categorical data
- Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding
- Set Block Decoding is a Language Model Inference Accelerator
- SSD-LM: Semi-autoregressive Simplex-based Diffusion Language Model for Text Generation and Modular Control
- ELF: Embedded Language Flows
- Adam: A Method for Stochastic Optimization
- Flow Map Language Models: One-step Language Modeling via Continuous Denoising
- TinyGSM: achieving >80% on GSM8k with small language models
- Decoupled Weight Decay Regularization
- Large Language Diffusion Models
- Discrete Flow Maps
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks