AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

FAFO: Lossy KV Cache Compression for Lossless Inference Acceleration via Draftless Fumble Decoding

Hoang Anh Duy Le, Shaochen Zhong, Yifan Lu, Yingtong Dou, Jiayi Yuan, Yu-Neng Chuang, Xiran Fan, Guanchu Wang, Yuzhong Chen, Xia Hu

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems

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
parallel draft-and-verify in one forward pass instead of two forward passes (ablation)FAFO-Stream · MT-Bench, Llama-2-7b-chatspeedup: 1.19× -> 1.91×Improved
parallel verification instead of two forwards (ablation)FAFO-Stream · MT-Bench, Llama-2-7b-chatacceptance length: 2.18 -> 2.29No clear effect
generate guesses on compressed KV cache instead of full cacheFAFO vs Lookahead · MT-Bench, Llama-2-7b-chatspeedup: 1.61× -> 1.91×Improved
increase number of n-gram guesses beyond about 10 (ablation)Lookahead · n-gram guesses sweepthroughputHurt
increase number of n-gram guesses (ablation)FAFO · n-gram guesses sweepthroughputImproved
cache n-grams under longer suffix prefixes (lookback window) (ablation)FAFO · n-gram guesses sweepacceptance lengthImproved
swap in Quest instead of StreamingLLM as compression method (ablation)FAFO · MT-Bench, Llama-2-7b-chatspeedup: 1.91× -> 1.32×Hurt
use FAFO instead of TriForceFAFO-Stream · MT-Bench, Llama-2-7b-chatspeedup: 0.21× -> 1.91×Improved
use FAFO instead of Lookahead on long-context LongBenchFAFO-Stream · LongBench Qasper, Llama-3.1-8Bspeedup: 0.85× -> 2.20×Improved
use FAFO-Stream vs Lookahead on Qwen2.5-32B GSM8KFAFO-Stream · GSM8K, Qwen2.5-32B-Instructspeedup: 1.30× vs 1.44× LookaheadHurt

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