# DFlare: Scaling Up Draft Capacity for Block Diffusion Speculative Decoding

arXiv: https://arxiv.org/abs/2606.02091 · HTML: https://mlsanity.com/paper/2606.02091/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Speculative decoding, one of 12 directions in Efficient LLM inference and serving.
- Builds on 29 of its 36 references that are in the map.
- First posted on arXiv 1 Jun 2026.

## Placed in

- Efficient LLM inference and serving / [Speculative decoding](https://mlsanity.com/inference/speculative-decoding/)

## Builds on (most cited first, 29 of 29)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [Structured Denoising Diffusion Models in Discrete State-Spaces](https://mlsanity.com/paper/2107.03006/) (2021)
- [Fast Inference from Transformers via Speculative Decoding](https://mlsanity.com/paper/2211.17192/) (2022)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [Diffusion-LM Improves Controllable Text Generation](https://mlsanity.com/paper/2205.14217/) (2022)
- [Accelerating Large Language Model Decoding with Speculative Sampling](https://mlsanity.com/paper/2302.01318/) (2023)
- [Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads](https://mlsanity.com/paper/2401.10774/) (2024)
- [Enhancing Chat Language Models by Scaling High-quality Instructional Conversations](https://mlsanity.com/paper/2305.14233/) (2023)
- [Simplified and Generalized Masked Diffusion for Discrete Data](https://mlsanity.com/paper/2406.04329/) (2024)
- [EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty](https://mlsanity.com/paper/2401.15077/) (2024)
- [Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding](https://mlsanity.com/paper/2505.22618/) (2025)
- [EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees](https://mlsanity.com/paper/2406.16858/) (2024)
- [Better & Faster Large Language Models via Multi-token Prediction](https://mlsanity.com/paper/2404.19737/) (2024)
- [EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test](https://mlsanity.com/paper/2503.01840/) (2025)
- [Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decoding](https://mlsanity.com/paper/2401.07851/) (2024)
- [DiffusionBERT: Improving Generative Masked Language Models with Diffusion Models](https://mlsanity.com/paper/2211.15029/) (2022)
- [DFlash: Block Diffusion for Flash Speculative Decoding](https://mlsanity.com/paper/2602.06036/) (2026)
- [Self-conditioned Embedding Diffusion for Text Generation](https://mlsanity.com/paper/2211.04236/) (2022)
- [NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model](https://mlsanity.com/paper/2508.14444/) (2025)
- [GliDe with a CaPE: A Low-Hassle Method to Accelerate Speculative Decoding](https://mlsanity.com/paper/2402.02082/) (2024)
- [Learning Harmonized Representations for Speculative Sampling](https://mlsanity.com/paper/2408.15766/) (2024)
- [Speculative Diffusion Decoding: Accelerating Language Generation through Diffusion](https://mlsanity.com/paper/2408.05636/) (2024)
- [DiffuSpec: Unlocking Diffusion Language Models for Speculative Decoding](https://mlsanity.com/paper/2510.02358/) (2025)
- [TiDAR: Think in Diffusion, Talk in Autoregression](https://mlsanity.com/paper/2511.08923/) (2025)
- [Bridging Draft Policy Misalignment: Group Tree Optimization for Speculative Decoding](https://mlsanity.com/paper/2509.22134/) (2025)
- [AngelSlim: A more accessible, comprehensive, and efficient toolkit for large model compression](https://mlsanity.com/paper/2602.21233/) (2026)

## Built on by (most cited first, 9 of 9)

- [D-cut: Adaptive Verification Depth Pruning for Batched Speculative Decoding](https://mlsanity.com/paper/2607.14647/) (2026)
- [AngelSpec: Towards Real-World High Performance Inference with Speculative Decoding](https://mlsanity.com/paper/2607.25852/) (2026)
- [DFlow: Enabling Verifier Information Flow in Block Diffusion Speculative Decoding](https://mlsanity.com/paper/2609.06498/) (2026)
- [Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding](https://mlsanity.com/paper/2609.36173/) (2026)
- [Fully Rotation-Equivariant Spectral-Spatial Learning for Multispectral Object Detection](https://mlsanity.com/paper/2607.05148/) (2026)
- [Ceiling-Clipped Acceptance Histograms Indicate Stranded Speed-up in Block-Diffusion Speculative Decoding](https://mlsanity.com/paper/2608.30427/) (2026)
- [LongSpark: Efficient speculative decoding with a fixed-cost parallel drafter](https://mlsanity.com/paper/2609.37029/) (2026)
- [DEdit: Iterative Draft Editing for Speculative Decoding](https://mlsanity.com/paper/2609.38510/) (2026)
- [DRelay: Global Draft Context for Prefix-Aware Parallel Speculative Decoding Repair](https://mlsanity.com/paper/2610.01439/) (2026)

## Abstract (arXiv)

Block diffusion speculative decoding accelerates LLM inference by predicting all tokens within a block simultaneously for the target model to verify in parallel. Predicting an entire block at once requires a sufficiently capable draft model and effective utilization of the target model's internal knowledge. However, the state-of-the-art method DFlash constrains all draft layers to share a single fused representation derived from only a few target layers, limiting per-layer expressiveness and hindering further scaling of draft capacity. In this paper, we present modelname, which flares out the narrow conditioning bottleneck of DFlash through a lightweight layer-wise fusion mechanism: each draft layer attends to its own learnable combination of a broad set of target layers at negligible overhead, simultaneously injecting richer target knowledge and providing every draft layer with a distinct input. This enhanced per-layer expressiveness enables scaling the draft model to deeper architectures with consistent gains. We further scale training data from 800K to 2.4M samples to fully exploit the enlarged capacity. On six benchmarks spanning mathematical reasoning, code generation, and conversation, modelname attains average wall-clock speedups of 5.52x on Qwen3-4B, 5.46x on Qwen3-8B, and 3.91x on GPT-OSS-20B, improving over DFlash by roughly 11%, 8%, and 5% respectively. Our code is available at https://github.com/Tencent/AngelSlim.