AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

TRACE: Trajectory-based Activation Change Estimation for Task-specific Data Selection

Ye He, Shangzhan Li, Yuxin Zhou, Qi Shi

AAAI 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Adaptive testing and experiments · 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 warmup stage (ablation)TRACE (LLaMA-3.1-8B, layer 15) · Avg BBH/MMLU/TyDiQAavg score: 65.93 -> 68.41Improved
add validation-set tuning step (ablation)TRACE (LLaMA-3.1-8B, layer 15) · Avg BBH/MMLU/TyDiQAavg score: 65.81 -> 68.41Improved
use official instruct model for selection instead of warmup model (ablation)TRACE (LLaMA-3.1-8B, layer 15) · Avg BBH/MMLU/TyDiQAavg score: 68.41 -> 66.42Hurt
aggregate TRACE scores over multiple warmup checkpoints (ablation)TRACE (LLaMA-3.1-8B) · Avg BBH/MMLU/TyDiQAavg score: 67.63 (1 epoch) vs 67.57/67.15/67.40No clear effect
select 5% with TRACE(kmeans) instead of full-data trainingLLaMA-3.1-8B · TyDiQAF1: 73.68 -> 76.50Improved
select 5% with TRACE instead of full dataOLMo-7B · Avg BBH/MMLU/TyDiQAavg score: 34.93 -> 35.10 (kmeans) / 34.69 (random)Mixed
random 5% selection vs full dataQwen-2.5-7B · Avg BBH/MMLU/TyDiQAavg score: 69.12 -> 69.76Improved
use TRACE instead of MoNALLaMA-3.1-8B · Avg BBH/MMLU/TyDiQAavg score: 67.20 -> 68.41Improved
use kmeans instead of random warmup data (ablation)TRACE (LLaMA-3.1-8B) · Avg BBH/MMLU/TyDiQAavg score: 67.42 -> 68.41Improved

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