Weaver builds conditional proposal trees from a factorized drafter’s top-K marginals and a rollback-free GDN tree-verify kernel, yielding 4.37× speedup over AR decoding.
hub
Fast- dllm v2: Efficient block-diffusion llm.arXiv preprint arXiv:2509.26328
29 Pith papers cite this work. Polarity classification is still indexing.
29
Pith papers citing it
hub tools
citation-role summary
background 3
baseline 1
citation-polarity summary
representative citing papers
Prefilling-dLLM partitions prefixes into chunks, caches KV representations, and applies sparse top-K selection during decoding to cut dLLM inference complexity to quadratic in decode length only.
MaskForge reaches 79.3% average attack success rate on five dLLMs by adaptively searching and accumulating structural attack patterns with a UCB bandit, improving 17.6% over baselines and transferring to 88.2% on AdvBench.
BASTION is a budget-aware speculative decoding framework with adaptive tree-structured block diffusion drafting that reports up to 6.61x speedup and 39% improvement over block-diffusion baselines.
TAD improves the accuracy-parallelism trade-off in diffusion LLMs via temporal-aware self-distillation that applies hard labels to soon-to-be-decoded tokens and soft supervision to future tokens.
DARE reuses up to 87% of attention activations in diffusion LLMs through KV caching and output reuse, delivering 1.2x per-layer latency gains with average performance drops of 1.2-2.0%.
Diffusion LLMs hallucinate more than autoregressive models and display distinct failure modes including premature termination, incomplete denoising, and context intrusion.
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.
Joint AR–diffusion training yields one tri-mode LM that switches AR, diffusion, and self-speculation, beating open AR/diffusion models on accuracy and tokens-per-forward.
CFT + on-policy distillation + RL converts an AR reasoning re-ranker into a block-diffusion model that recovers near-AR accuracy at 2.4–3.5× decode throughput on Amazon Beauty.
DiLaServe improves SLO attainment for diffusion language models by up to 56.6 percentage points and reduces latency by up to 46% with less than 1% accuracy drop via deadline-aware scheduling and dynamic reconfiguration.
R2LM combines causal attention with a reverse Mamba SSM sidecar to supply right-side context in dLLMs, claiming 2.4x-12.9x throughput gains over bidirectional dLLMs and 1.9x-2.9x over AR baselines while matching or exceeding quality.
JetSpec trains a causal draft head to produce branch-consistent trees aligned with target autoregressive scores, achieving up to 9.64x speedup on MATH-500 and outperforming prior SD baselines on Qwen3 models.
On-policy distillation from a frozen autoregressive teacher to a bidirectional student eliminates train-inference mismatch and enables data-efficient ARLM-to-DLM conversion.
Introduces TSPD with a trajectory-feature controller and training-free CE to reduce denoising steps in dLLMs while aiming to preserve quality.
Block-diffusion decoder with prior-calibrated scoring and early stopping produces streaming zero-shot TTS at quality comparable to AR and NAR baselines with lower real-time factor.
Learned Relay Representations add a differentiable per-token channel to masked diffusion models so they can propagate latent information across iterative denoising steps, yielding better coding performance and up to 32% lower latency on Fast-dLLM v2 than standard supervised finetuning.
FlexDraft is a lossless speculative decoding framework that adapts to batch sizes via attention tuning on final layers, MLP-based bonus calibration, and dynamic parallel/sequential decoding.
Orthrus unifies autoregressive LLMs and diffusion models via shared KV cache and consensus to enable up to 7.8x parallel token generation speedup with O(1) memory overhead and lossless results.
ML-Bench is a multilingual safety benchmark derived from actual regional laws and regulations, paired with ML-Guard guardrail models that outperform 11 baselines on existing and new benchmarks.
Progressive supervised block merging plus stage-wise intra-diffusion distillation converts Qwen3-VL into a large-block dVLM with claimed SOTA open-dVLM scores and up to 3× decoding speedup on ≤4.4M data.
DLMs exhibit lower n-gram entropy, higher semantic coherence, and higher semantic diversity than ARMs, primarily due to bidirectional context and remasking decoding strategies.
Three expert-activation strategies (delayed caching, speculative exploration, limited activation) accelerate MoE diffusion LLMs by up to 2.2x with minimal quality change.
AsyncVLA adds asynchronous flow matching and a confidence rater to VLA models so they can generate actions on flexible schedules and selectively refine low-confidence tokens before execution.
citing papers explorer
-
Trees from Marginals: Autoregressive drafting with factorized priors
Weaver builds conditional proposal trees from a factorized drafter’s top-K marginals and a rollback-free GDN tree-verify kernel, yielding 4.37× speedup over AR decoding.
-
Prefilling-dLLM: Predictive Prefilling for Long-Context Inference in Diffusion Language Models
Prefilling-dLLM partitions prefixes into chunks, caches KV representations, and applies sparse top-K selection during decoding to cut dLLM inference complexity to quadratic in decode length only.
-
MaskForge: Structure-Aware Adaptive Attacks for Jailbreaking Diffusion Large Language Models
MaskForge reaches 79.3% average attack success rate on five dLLMs by adaptively searching and accumulating structural attack patterns with a UCB bandit, improving 17.6% over baselines and transferring to 88.2% on AdvBench.
-
Bastion: Budget-Aware Speculative Decoding with Tree-structured Block Diffusion Drafting
BASTION is a budget-aware speculative decoding framework with adaptive tree-structured block diffusion drafting that reports up to 6.61x speedup and 39% improvement over block-diffusion baselines.
-
TAD: Temporal-Aware Trajectory Self-Distillation for Fast and Accurate Diffusion LLM
TAD improves the accuracy-parallelism trade-off in diffusion LLMs via temporal-aware self-distillation that applies hard labels to soon-to-be-decoded tokens and soft supervision to future tokens.
-
DARE: Diffusion Language Model Activation Reuse for Efficient Inference
DARE reuses up to 87% of attention activations in diffusion LLMs through KV caching and output reuse, delivering 1.2x per-layer latency gains with average performance drops of 1.2-2.0%.
-
Lost in Diffusion: Uncovering Hallucination Patterns and Failure Modes in Diffusion Large Language Models
Diffusion LLMs hallucinate more than autoregressive models and display distinct failure modes including premature termination, incomplete denoising, and context intrusion.
-
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.
-
Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding
Joint AR–diffusion training yields one tri-mode LM that switches AR, diffusion, and self-speculation, beating open AR/diffusion models on accuracy and tokens-per-forward.
-
Diffusion-GR2: Diffusion Generative Reasoning Re-ranker
CFT + on-policy distillation + RL converts an AR reasoning re-ranker into a block-diffusion model that recovers near-AR accuracy at 2.4–3.5× decode throughput on Amazon Beauty.
-
DiLaServe: High SLO Attainment Serving for Diffusion Language Models
DiLaServe improves SLO attainment for diffusion language models by up to 56.6 percentage points and reduces latency by up to 46% with less than 1% accuracy drop via deadline-aware scheduling and dynamic reconfiguration.
-
Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation
R2LM combines causal attention with a reverse Mamba SSM sidecar to supply right-side context in dLLMs, claiming 2.4x-12.9x throughput gains over bidirectional dLLMs and 1.9x-2.9x over AR baselines while matching or exceeding quality.
-
JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting
JetSpec trains a causal draft head to produce branch-consistent trees aligned with target autoregressive scores, achieving up to 9.64x speedup on MATH-500 and outperforming prior SD baselines on Qwen3 models.
-
Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation
On-policy distillation from a frozen autoregressive teacher to a bidirectional student eliminates train-inference mismatch and enables data-efficient ARLM-to-DLM conversion.
-
Efficient Diffusion LLMs via Temporal-Spatial Parallel Decoding and Confidence Extrapolation
Introduces TSPD with a trajectory-feature controller and training-free CE to reduce denoising steps in dLLMs while aiming to preserve quality.
-
Chatterbox-Flash: Prior-Calibrated Block Diffusion for Streaming Zero-Shot TTS
Block-diffusion decoder with prior-calibrated scoring and early stopping produces streaming zero-shot TTS at quality comparable to AR and NAR baselines with lower real-time factor.
-
Learned Relay Representations for Forward-Thinking Discrete Diffusion Models
Learned Relay Representations add a differentiable per-token channel to masked diffusion models so they can propagate latent information across iterative denoising steps, yielding better coding performance and up to 32% lower latency on Fast-dLLM v2 than standard supervised finetuning.
-
FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided Calibration
FlexDraft is a lossless speculative decoding framework that adapts to batch sizes via attention tuning on final layers, MLP-based bonus calibration, and dynamic parallel/sequential decoding.
-
Orthrus: Memory-Efficient Parallel Token Generation via Dual-View Diffusion
Orthrus unifies autoregressive LLMs and diffusion models via shared KV cache and consensus to enable up to 7.8x parallel token generation speedup with O(1) memory overhead and lossless results.
-
ML-Bench&Guard: Policy-Grounded Multilingual Safety Benchmark and Guardrail for Large Language Models
ML-Bench is a multilingual safety benchmark derived from actual regional laws and regulations, paired with ML-Guard guardrail models that outperform 11 baselines on existing and new benchmarks.
-
BARD: Bridging AutoRegressive and Diffusion Vision-Language Models Via Highly Efficient Progressive Block Merging and Stage-Wise Distillation
Progressive supervised block merging plus stage-wise intra-diffusion distillation converts Qwen3-VL into a large-block dVLM with claimed SOTA open-dVLM scores and up to 3× decoding speedup on ≤4.4M data.
-
Differences in Text Generated by Diffusion and Autoregressive Language Models
DLMs exhibit lower n-gram entropy, higher semantic coherence, and higher semantic diversity than ARMs, primarily due to bidirectional context and remasking decoding strategies.
-
TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration
Three expert-activation strategies (delayed caching, speculative exploration, limited activation) accelerate MoE diffusion LLMs by up to 2.2x with minimal quality change.
-
AsyncVLA: Asynchronous Flow Matching for Vision-Language-Action Models
AsyncVLA adds asynchronous flow matching and a confidence rater to VLA models so they can generate actions on flexible schedules and selectively refine low-confidence tokens before execution.
-
BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding
BlockPilot is an instance-adaptive policy that predicts optimal block size from the prefilling representation for diffusion speculative decoding, reporting 5.92 acceptance length and 4.20x speedup on Qwen3-4B.
-
WaveFilter: Enhancing the Long-Context Capability of Diffusion LLMs via Wavelet-Guided KV Cache Filtering
WaveFilter applies wavelet decomposition to filter critical tokens for sparse KV caching, improving long-context performance of diffusion LLMs as a plug-and-play addition to existing methods.
-
Reinforcement Learning from Denoising Feedback
RLDF is a new RL paradigm for diffusion language models that optimizes toward clipped clean states with weighted timestep sampling and reports substantial gains on reasoning benchmarks for LLaDA and Dream.
-
Multi-Token Residual Prediction
MRP predicts logit residuals between adjacent denoising steps in DLMs from backbone hidden states to support efficient multi-token denoising, yielding up to 1.4x lossless speedup or 22.6-point accuracy gains on code and math tasks.
-
ECHO: Efficient Chest X-ray Report Generation with One-step Block Diffusion
ECHO introduces one-step block diffusion via Direct Conditional Distillation and Response-Asymmetric Diffusion to generate chest X-ray reports faster than autoregressive models while improving clinical metrics.