mlsanityData to 7 Oct 2026

CORA-Diff: Confidence-Oriented Residual Acceptance for Efficient Diffusion Language Model Inference

arXiv 2608.11235 ↗First posted 31 Jul 2026cs.AIcited by 0

On the map: Diffusion models → Discrete and text diffusion

Abstract (arXiv)

Diffusion language models (DLMs) update many tokens in parallel, yet practical decoders often use a fixed denoising horizon. Many predictions stabilize early, but blockwise decoding continues until all positions are resolved, causing repeated dense forward passes. Existing accelerators often rely on learned filters, modified scores, dependency models, or cache-specific mechanisms. We ask whether native trajectory signals can identify residual positions likely to match the deterministic dense endpoint. We propose CORA-Diff, a training-free method that preserves the original transfer rule and applies confidence-and-persistence gating only to positions that rule leaves unresolved. Accepted tokens remain visible as context, and the block terminates once all positions are resolved. This requires no backbone change, learned acceptance model, or logit modification. Our theory explains why high-confidence, persistent predictions are more likely to match the fixed-horizon dense endpoint, and paired post-intervention trajectories provide direct empirical support. We select one operating point on a separate GSM8K calibration subset and freeze it for all evaluations. Under a matched Learn2PD-style LLaDA protocol, CORA-Diff has the lowest measured runtime in all eight task-length settings. Task scores match or exceed dense decoding in five settings, and the largest observed drop is 1.22 points. Its incremental speedups over EOS-aware dense decoding are 2.70x and 3.32x on GSM8K and HumanEval. It also reaches 13.14x under the fixed-horizon 1024/1024 mechanism-isolation protocol and transfers to Dream without retuning at 3.18x-3.53x. These results show that native confidence and persistence enable reliable residual acceptance, reducing repeated denoising computation while preserving task quality.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Discrete and text diffusion, one of 14 directions in Diffusion models. Builds on 24 of its 33 references that are in the map. First posted on arXiv 31 Jul 2026.

Builds on

24 of 33 references.

Closest ancestors · 9

Same direction, newest first

  1. 2026Attention-Discounted Adaptive Sampler for Masked Diffusion Language Models9 Jun
  2. 2026DMax: Aggressive Parallel Decoding for dLLMs9 Apr
  3. 2025WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference28 Dec
  4. 2025KLASS: KL-Guided Fast Inference in Masked Diffusion Models7 Nov
  5. 2025FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion27 May
  6. 2024Scaling Diffusion Language Models via Adaptation from Autoregressive Models23 Oct
  7. 2024Simplified and Generalized Masked Diffusion for Discrete Data6 Jun
  8. 2024Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data6 Jun
  9. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution25 Oct

Other references · 15

  1. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  2. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  3. 2022A Continuous Time Framework for Discrete Denoising Modelscited by 437
  4. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decodingcited by 404
  5. 2024LayerSkip: Enabling Early Exit Inference and Self-Speculative Decodingcited by 302
  6. 2020DynaBERT: Dynamic BERT with Adaptive Width and Depthcited by 246
  7. 2025dParallel: Learnable Parallel Decoding for dLLMscited by 59
  8. 2025Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principlescited by 43
  9. 2025AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Sizecited by 30
  10. 2025Learning to Parallel: Accelerating Diffusion Large Language Models via Learnable Parallel Decodingcited by 26
  11. 2026DAPD: Dependency-Aware Parallel Decoding via Attention for Diffusion LLMscited by 13
  12. 2026DAWN: Dependency-Aware Fast Inference for Diffusion LLMscited by 12
3 more
  1. 2026FOCUS: DLLMs Know How to Tame Their Compute Boundcited by 3
  2. 2026TIDE: Efficient and Lossless MoE Diffusion LLM Inference with I/O-aware Expert Offloadcited by 2
  3. 2026Efficient Diffusion LLMs via Temporal-Spatial Parallel Decoding and Confidence Extrapolationcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026CAI-DLLM: Convergence Aware Inference for Diffusion Language Models
  2. 2026CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models
  3. 2025Beyond Confidence: Adaptive and Coherent Decoding for Diffusion Language Models
  4. 2026CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models
  5. 2026LESS Is More: Mutual-Stability Sampling for Diffusion Language Models
  6. 2025Sequential Diffusion Language Models
  7. 2026AsyncLane: Decoupling Refinement from Advancement in Diffusion Language Model Decoding
  8. 2026Self-Distilled Trajectory-Aware Boltzmann Modeling: Bridging the Training-Inference Discrepancy in Diffusion Language Models