AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Scheduling LLM Inference with Uncertainty-Aware Output Length Predictions

Haoyu Zheng, Yongqiang Zhang, Fangcheng Fu, Xiaokai Zhou, Hao Luo, Hongchao Zhu, Yuanyuan Zhu, Hao Wang, Xiao Yan, Jiawei Jiang

ICML 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Priors and functional variational uncertainty · 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
model output length as a log-t distribution and schedule by Tail Inflated Expectation instead of point-estimate SJF (LTR)LLM request scheduler · LMSYS-Chat-1M online, 100 RPSaverage per-token latency: 2.31x reduction over LTR; 4.73x over FCFSImproved
use TIE instead of LTR for offline synthetic data generationLLM request scheduler · Alpaca SDG, 8Btime for 3k samples: 139.53 -> 98.12Improved
replace log-t with log-normal distribution (ablation)TIE · LMSYS-Chat-1M onlineaverage per-token latency: 0.67 -> 1.63Hurt
learn dynamic nu in log-t instead of fixing nu=3.5 (ablation)TIE · LMSYS-Chat-1M onlineaverage per-token latency: 0.67 -> 0.69No clear effect
add CVaR tail term to expected length score (ablation)TIE · LMSYS-Chat-1M onlineaverage per-token latency: 0.75 -> 0.72 (beta=0.1)Improved
use adaptive beta instead of fixed beta (ablation)TIE · LMSYS-Chat-1M onlineaverage per-token latency: 0.71 (beta=0.3) -> 0.67Improved
use CVaR tail term in SDG (ablation)TIE · Alpaca SDGthroughput within 3 min: 4257 -> 4762Improved
train predictor on Alpaca instead of LMSYS-Chat-1M (ablation)TIE · Alpaca SDG, 8Btime for 3k samples: 98.12 -> 95.08Improved
transfer TIE to different datasets and a 70B model without retrainingTIE · ShareGPT/Alpaca, 70Bper-token latency: 3.77x over FCFS on 70BImproved
risk-adaptive beta under rising load from 30 to 100 RPSTIE · LMSYS-Chat-1M onlineper-token latency growth: 7.42x/8.55x/6.17x -> 3.68xImproved

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