AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Three Forward, One Backward: Memory-Efficient Full-Rank Fine-Tuning of Large Models via Extra Forward Passes

Jia Zhang, Yu Bai, Hualin Zhang, Tianshuo Chen, Zhaogeng Liu, zhiqiang xu, Yi Chang, Bin Gu

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning

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 zeroth-order full-weight updates to LoRA (alternating)LoRA · SST-2 (OPT-1.3B)accuracy: 92.2 -> 92.4Improved
add zeroth-order full-weight updates to LoRA (alternating)LoRA · BoolQ (OPT-6.7B)accuracy: 49.4 -> 65.1Improved
add zeroth-order full-weight updates to LoRALoRA · RTE (OPT-1.3B)accuracy: 61.1 -> 59.4Hurt
add LoRA updates to MeZOMeZO · SST-2 (OPT-1.3B)accuracy: 58.1 -> 92.4Improved
add LoRA updates to MeZOMeZO · SST-2 (RoBERTa-large k=16)accuracy: 90.5 -> 92.8Improved
alternate low-rank and ZO updates (full LMAO) vs components alone (w/o LoRA or w/o MeZO) (ablation)LMAO · OPT-1.3B six tasksaccuracyImproved
increase training data from k=16 to k=512 per classLMAO · RoBERTa-large (6 tasks)accuracy: gap to FT narrows; SNLI 87.9 vs FT 88.7Mixed
vary LoRA rank r and scaling alpha (ablation)LMAO · SST-2 (RoBERTa-large)accuracyNo clear effect
add LoRA update on top of ZO step (alternating) memoryLMAO · peak GPU memory BoolQ OPT-1.3BGB: 38.48 -> 23.08Improved
use LMAO versus LoRA/MeZO on OPT-6.7B WICLMAO · WIC (OPT-6.7B)accuracy: 52.9 -> 50.0Hurt

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