# PACE-dLLM: Elastic Block Decoding via Confidence Cliff Estimation for Diffusion Language Models

arXiv: https://arxiv.org/abs/2609.26249 · HTML: https://mlsanity.com/paper/2609.26249/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 7 of its 20 references that are in the map.
- First posted on arXiv 13 Aug 2026.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)
- Efficient LLM inference and serving / [Diffusion LLM and MoE inference](https://mlsanity.com/inference/diffusion-llm-and-moe-inference/)

## Builds on (most cited first, 7 of 7)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [Dream 7B: Diffusion Large Language Models](https://mlsanity.com/paper/2508.15487/) (2025)
- [When to Commit? Towards Variable-Size Self-Contained Blocks for Discrete Diffusion Language Models](https://mlsanity.com/paper/2604.23994/) (2026)
- [DepCap: Adaptive Block-Wise Parallel Decoding for Efficient Diffusion LM Inference](https://mlsanity.com/paper/2604.15750/) (2026)

## Abstract (arXiv)

Diffusion language models (dLLMs), such as LLaDA and Dream, have become competitive with autoregressive (AR) LLMs in generation quality while supporting native parallel decoding. A standard acceleration strategy is block-wise decoding, where each forward pass predicts a block of length B and commits high-confidence tokens. However, B couples two distinct decisions: the look-ahead horizon and the number of tokens to commit. Existing accelerators address this limitation through indirect heuristics, such as volatility tracking, delimiter detection, and learned scoring. In contrast, we show that the required information is already encoded in the model's own per-step confidence: in-window confidence typically follows a context-dependent cliff, whose saturation point directly identifies the appropriate look-ahead horizon. We propose PACE-dLLM, which fits this parametric cliff in closed form at each step, sets the next horizon by its saturation point, and uses an independent confidence threshold for token commitment. Under a saturated-yield abstraction, we show that the cliff-anchored horizon is the smallest horizon attaining maximal useful per-pass yield: fixed horizons that undershoot it incur a worse asymptotic NFE rate, while overshooting adds no useful yield. On four reasoning and code benchmarks, PACE-dLLM achieves the best average accuracy on both open-source dLLM backbones, with average wall-clock speedups of 5.23x on LLaDA and 3.06x on Dream (up to 8.52x on math) over the unaccelerated semi-AR baseline, advancing the quality-throughput Pareto frontier.