AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Dustin: Draft-Augmented Sparse Verification for Efficient Long-Context Generation with Speculative Decoding

WenHung Lee, Jian-Jia Chen, Xiaolin Lin, Pei-Shuo Wang, Chi-Chih Chang, Chun-Che Yang, Wei-Chen Wang, Hanrui Wang, Ning-Chi Huang, Grace Li Zhang, Kai-Chiang Wu

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Long-context KV and memory attention · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add draft-based lookahead attention to target-historical importance estimation (ablation)Dustin-T · TriviaQAaccuracy (%): 77.58% -> 83.63%Improved
add target-historical attention to draft-lookahead importance estimation (ablation)Dustin-D · TriviaQAaccuracy (%): 80.51% -> 83.63%Improved
add draft-based lookahead attention to target-historical importance estimation (ablation)Dustin-T · MultiNewsscore (%): 23.95% -> 25.19%Improved
add target-historical attention to draft-lookahead importance estimation (ablation)Dustin-D · MultiNewsscore (%): 24.84% -> 25.19%Improved
use Dustin hybrid sparse verification instead of Quest at tight 128-token KV budgetQwen2.5-72B · LongBench avg (KV budget 128)score: 37.01 -> 52.41Improved
use Dustin instead of SnapKV at 128-token KV budgetLlama-3.3-70B · LongBench avg (KV budget 128)score: 50.32 -> 53.81Improved
increase KV budget from 128 to 512 (ablation)Dustin on Qwen2.5-72B · LongBench avgscore: 52.41 -> 55.23Improved
increase context length from 8K to 32K (batch 8)Dustin on Qwen2.5-72B · PG-19 decode speedupspeedup vs Vanilla: 3.01x -> 6.61xImproved
increase context length from 8K to 32K (batch 16)Dustin on Qwen2.5-72B · PG-19 decode speedupspeedup vs Vanilla: 4.76x -> 9.17xImproved
use Dustin instead of MagicDec at 32K context, batch 8Llama-3.3-70B · PG-19 decode speedupspeedup vs Vanilla: 2.05x -> 6.28xImproved
restrict importance estimation to selected retrieval heads (ablation)Dustin · importance-estimation overheadfraction of full hybrid computation: about 0.8%Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy