DMax uses On-Policy Uniform Training and Soft Parallel Decoding to enable aggressive parallelism in dLLMs, raising TPF on GSM8K from 2.04 to 5.47 and on MBPP from 2.71 to 5.86 while preserving accuracy.
Title resolution pending
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
CoDiLA adds a compact auxiliary AR model on diffusion latents to enforce local sequential validity during parallel token sampling in discrete diffusion language models.
Early and late denoising steps in masked diffusion LMs are robust to smaller-model replacement, enabling 17% FLOPs reduction with modest generative quality loss.
PulseCol introduces periodically refreshed column-sparse attention to achieve up to 1.95x speedup over FlashAttention in diffusion LLMs with maintained model quality.
Position-preserving MASK token compression reduces redundancy in diffusion LLMs to accelerate parallel decoding and enable context folding for longer sequences.
STDec raises dLLM decoding speed by up to 14x on benchmarks like MBPP by using observed spatio-temporal stability to create dynamic, token-specific confidence thresholds while preserving task performance.
ETS performs training-free RL alignment for language models by energy-guided test-time scaling with Monte Carlo energy estimation and importance sampling acceleration.
DLMs show early answer convergence allowing Prophet to cut decoding steps by up to 3.4x on LLaDA-8B and Dream-7B while keeping output quality.
citing papers explorer
-
DMax: Aggressive Parallel Decoding for dLLMs
DMax uses On-Policy Uniform Training and Soft Parallel Decoding to enable aggressive parallelism in dLLMs, raising TPF on GSM8K from 2.04 to 5.47 and on MBPP from 2.71 to 5.86 while preserving accuracy.
-
Locally Coherent Parallel Decoding in Diffusion Language Models
CoDiLA adds a compact auxiliary AR model on diffusion latents to enforce local sequential validity during parallel token sampling in discrete diffusion language models.
-
Not All Denoising Steps Are Equal: Model Scheduling for Faster Masked Diffusion Language Models
Early and late denoising steps in masked diffusion LMs are robust to smaller-model replacement, enabling 17% FLOPs reduction with modest generative quality loss.
-
PulseCol: Periodically Refreshed Column-Sparse Attention for Accelerating Diffusion Language Models
PulseCol introduces periodically refreshed column-sparse attention to achieve up to 1.95x speedup over FlashAttention in diffusion LLMs with maintained model quality.
-
Elastic-dLLM: Position Preserving Context Compression and Augmentation of Diffusion LLMs
Position-preserving MASK token compression reduces redundancy in diffusion LLMs to accelerate parallel decoding and enable context folding for longer sequences.
-
STDec: Spatio-Temporal Stability Guided Decoding for dLLMs
STDec raises dLLM decoding speed by up to 14x on benchmarks like MBPP by using observed spatio-temporal stability to create dynamic, token-specific confidence thresholds while preserving task performance.
-
ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment
ETS performs training-free RL alignment for language models by energy-guided test-time scaling with Monte Carlo energy estimation and importance sampling acceleration.
-
Diffusion Language Models Know the Answer Before Decoding
DLMs show early answer convergence allowing Prophet to cut decoding steps by up to 3.4x on LLaDA-8B and Dream-7B while keeping output quality.
- $R^2$-dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction