AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Revisiting the Role of Pretrained Weights in Model Merging: On Near-Optimality within the Core Subspace

Wenju Sun, Qingyong Li, Tiancheng Li, Yangliao Geng, Boyang Li

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging

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 CSP projectionTask Arithmetic · 8 vision tasks, ViT-B/32average accuracy: 69.1 -> 72.7Improved
add CSP projectionTSV · 8 vision tasks, ViT-B/32average accuracy: 84.1 -> 85.1Improved
add CSP projectionISO-CTS · 8 vision tasks, ViT-B/32average accuracy: 84.3 -> 84.7Improved
add CSP projectionTask Arithmetic · 8 vision tasks, ViT-L/14average accuracy: 84.5 -> 86.2Improved
add CSP projectionTSV · 20 vision tasks, ViT-B/32average accuracy: 73.20 -> 75.06Improved
add CSP projectionTask Arithmetic · OKVQA (merged BLIP)accuracy: 17.71 -> 24.67Improved
add CSP projectionTask Arithmetic · TextVQA (merged BLIP)accuracy: 0.49 -> 10.28Improved
add CSP projectionTask Arithmetic · COCO Caption (merged BLIP)CIDEr: 0.86 -> 0.83Hurt
add CSP projectionTSV · COCO Caption (merged BLIP)CIDEr: 0.84 -> 0.81Hurt
add CSP projectionTask Arithmetic · Banking77 (Qwen3-0.6B)accuracy: 74.67% -> 74.16%Hurt
add CSP projectionTask Arithmetic and TSV · HotpotQA (Qwen3-8B)NDCG@10: 0.64 -> 0.65 (TA); 0.75 -> 0.76 (TSV)Improved
add CSP projectionKnOTS · 7 vision tasks (LoRA merging)average accuracy: 54.51 -> 56.38Improved

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