AI Paper Atlas

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

From Memorization to Parameter Interference: How Overtraining Experts Harms Model Merging

Stefan Horoi, Guy Wolf, Eugene Belilovsky, Gintare Karolina Dziugaite

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data · 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
fine-tune experts 256 steps instead of 2048 (ablation)Task Arithmetic merge of FFT ViT-B-32 · 8 vision tasksmerged accuracy (%): 69.8 -> 72.7Improved
fine-tune LoRA experts 128 steps instead of 2048 (ablation)TIES merge of LoRA ViT-B-32 · 8 vision tasksmerged accuracy (%): 50.7 -> 66.9Improved
train experts longer (256 to 2048 steps) (ablation)FFT ViT-B-32 experts · 8 vision tasksexpert accuracy (%): 88.4% -> 90.0%Improved
train experts longer (256 to 2048 steps) (ablation)FFT T5-Base experts · 7 NLP tasksexpert accuracy (%): 78.2% -> 81.9%Improved
train experts longer (256 to 2048 steps) (ablation)Task Arithmetic, TIES, DARE merges of FFT ViT-B-32 · 8 vision tasksmerged accuracy (%): about 3% dropHurt
train experts longer (ablation)Average merge of T5-Base · 7 NLP tasksmerged accuracy (%): peak 65.2 maintainedNo clear effect
use early stopping on expert trainingTIES merge of T5-Base · 7 NLP tasksmerged accuracy (%): 72.2 -> 76.9Improved
use early stopping on expert trainingAverage merge of T5-Base · 7 NLP tasksmerged accuracy (%): 65.2 -> 63.4Hurt
use early stopping on LoRA ViT expertsTask Arithmetic merge · 8 vision tasksmerged accuracy (%): 61.1 -> 68.0Improved
use early stopping on FFT ViT expertsAverage merge · 8 vision tasksmerged accuracy (%): 65.9 -> 64.5Hurt
remove hardest examples (top 1-10% EL2N) from expert training (ablation)merges of ViT-B-32 experts · 8 vision tasksmerged accuracyHurt
increase LoRA rank (ablation)LoRA experts merging · vision/NLPmerged accuracyImproved

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