AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

TuneAhead: Predicting Fine-tuning Performance Before Training Begins

Yuxiang Luo, Haonan Long, Chen Wang, Qiqi Duan, Xiaotian Lin, Yanwei Xu, Yuyu Luo, Weikai Yang, Nan Tang

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning 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
add dynamic probe features to static-only features (ablation)TuneAhead · MMLU performance predictionRMSE (pp): 3.50 -> 1.47Improved
add static dataset features to dynamic-only features (ablation)TuneAhead · MMLU performance predictionRMSE (pp): 3.38 -> 1.47Improved
add static features to dynamic-only predictor on runs where only dynamic succeeded (ablation)TuneAhead · MMLU (rescued-by-dynamic bucket)Acc@2pp: 38.7 -> 75.0Improved
add dynamic features to static-only predictor on runs where only static succeeded (ablation)TuneAhead · MMLU (rescued-by-static bucket)Acc@2pp: 37.3 -> 100.0Improved
combine static and dynamic features when both subset predictors failed (ablation)TuneAhead · MMLU (both-wrong bucket)Acc@2pp: 12.5 -> 70.2Improved
add a 100-step probe (dynamic features) versus 0 steps (ablation)TuneAhead · MMLU performance predictionRMSE (pp): 3.50 -> 1.47Improved
extend probe from 100 to 200 steps (ablation)TuneAhead · MMLU performance predictionAcc@2pp: 82.5 -> 82.8No clear effect
extend probe from 100 to 200 steps (ablation)TuneAhead · MMLU performance predictionprobe time (min): 7.76 -> 12.11Hurt
use TuneAhead instead of ProxyLMTuneAhead · MMLU performance predictionRMSE (pp): 2.11 -> 1.47Improved
use TuneAhead instead of ProxyLM on Llama-3-8B-InstructTuneAhead · MMLU (Llama-3-8B-Instruct)RMSE (pp): 6.76 -> 5.02Improved
use TuneAhead instead of ProxyLM on TruthfulQATuneAhead · TruthfulQA MC2 performance predictionRMSE (pp): 2.54 -> 2.17Improved
apply diagnosis-guided fixes (lower lr, SemDeDup, outlier removal) to a predicted-failure runTuneAhead diagnosis · MMLU single runMMLU accuracy: 20.2 -> 48.7Improved

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