AI Paper Atlas

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

Evaluating Parameter Efficient Methods for RLVR

Qingyu Yin, Yulun Wu, Zhennan Shen, Sunbowen Lee, Zhilin Wang, Yanshu Li, Chak Tou Leong, Jiale Kang, Jinjin Gu

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
use DoRA instead of LoRADeepSeek-R1-Distill-Qwen-1.5B RLVR · Math suite avgaverage accuracy: 42.5 -> 46.6Improved
use AdaLoRA instead of LoRADeepSeek-R1-Distill-Qwen-1.5B RLVR · Math suite avgaverage accuracy: 42.5 -> 44.2Improved
use MiSS instead of LoRADeepSeek-R1-Distill-Qwen-1.5B RLVR · Math suite avgaverage accuracy: 42.5 -> 43.4Improved
use LoRA+ learning-rate ratioLoRA · Math suite avgaverage accuracy: 42.5 -> 43.9Improved
use PiSSA principal-component initializationLoRA · Math suite avgaverage accuracy: 42.5 -> 0.2Hurt
use MiLoRA minor-component initializationLoRA · Math suite avgaverage accuracy: 42.5 -> 18.0Hurt
freeze matrix A (LoRA-FA)LoRA · Math suite avgaverage accuracy: 42.5 -> 43.0No clear effect
use VeRA frozen-projection scaling vectorsLoRA · Math suite avgaverage accuracy: 42.5 -> 40.7Hurt
use IA3 activation scalingLoRA · Math suite avgaverage accuracy: 42.5 -> 22.3Hurt
reduce LoRA rank to 1 (ablation)LoRA · Math suite avgaverage accuracy: 42.5 -> 40.5Hurt
raise LoRA rank to 16 (ablation)LoRA · Math suite avgaverage accuracy: 42.5 -> 43.9Improved
reduce batch size from 128 to 32 (ablation)LoRA · Math suite avgaverage accuracy: 42.5 -> 43.0Improved

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