AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

PiCa: Parameter-Efficient Fine-Tuning with Column Space Projection

Junseo Hwang, Wonguk Cho, Taesup Kim

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning · 6 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
project gradients onto the column space of pre-trained weights instead of a random space (ablation)PiCa · Commonsense Reasoning (Gemma-2B)avg accuracy: 63.18 -> 67.60Improved
add weight-sharing across layers (rank 128, 5.1M params) versus no sharing (rank 16, 35.8M params) (ablation)PiCa · Commonsense Reasoning (Gemma-7B)accuracy: 35.8M -> 5.1M trainable parametersNo clear effect
add weight-sharing under varying ranks (ablation)PiCa · GSM-8K (Gemma-2B)accuracyImproved
increase PiCa rank from 32 to 256PiCa · GSM-8K (Gemma-2B)accuracy: 41.32 -> 52.77Improved
increase PiCa rank from 32 to 256PiCa · GSM-8K (LLaMA-3-8B)accuracy: 73.54 -> 76.12Improved
increase PiCa rank from 16 to 128PiCa · Commonsense Reasoning (Gemma-7B)avg accuracy: 82.83 -> 84.47Improved
use PiCa r=256 instead of SVFT-RSVFT-R · GSM-8K (Gemma-2B)accuracy: 50.03 -> 52.77Improved
use PiCa r=32 instead of SVFT-PSVFT-P · GSM-8K (Gemma-2B)accuracy: 40.34 -> 41.32Improved
use PiCa r=128 instead of LoRA r=32LoRA · Commonsense Reasoning (Gemma-7B)avg accuracy: 83.69 -> 84.47Improved
use PiCa r=16 instead of SVFT-PSVFT-P · Commonsense Reasoning (Gemma-7B)avg accuracy: 80.85 -> 82.83Improved
use PiCa instead of BOFTBOFT · GLUE (DeBERTaV3-base)avg score: 89.89 -> 89.69No clear effect
use PiCa r=16 instead of LoRA r=8LoRA · GLUE (DeBERTaV3-base)avg score: 88.50 -> 89.69Improved

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