AI Paper Atlas

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

Gradient Transformer: Learning to Generate Updates for LLMs

Binh-Nguyen Nguyen, Khang Tran, Hai Phan, Issa Khalil

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems · 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
Generate LLM update vectors from TinyLM updates with Grad-Transformer instead of fine-tuned TinyLM aloneQwen2.5-3B to 7B · AQuA-RATaccuracy: 48.43 -> 61.02Improved
Use Grad-Transformer instead of Conf baselineQwen2.5-3B to 7B · Six datasetsaverage PGR: 58.94 -> 91.88Improved
Use Grad-Transformer instead of Conf baselineQwen2.5-3B to 7B · GSM8Kaccuracy: 74.30 -> 73.59Hurt
Use Grad-Transformer instead of Conf baselineQwen2.5-3B to 7B · CommonsenseQAaccuracy: 83.46 -> 83.21Hurt
Aggregate five clients' TinyLM updatesGrad-Transformer · Multi-client PGRPGR: SAMSum: Conf 98.24 vs ours 78.53; AQuA-RAT 34.70 -> 85.15Mixed
Apply DP-SGD with epsilon 0.1 to client TinyLM trainingGrad-Transformer · DROPaccuracy: TinyLM 55.59 -> 44.70; Grad-Transformer 62.35Mixed
Shrink TinyLM from 3B to 0.5B (ablation)Grad-Transformer to 7B LLM · AQuA-RATaccuracy: 63.57 -> 63.28No clear effect
Scale LLM from 7B to 14B (ablation)Grad-Transformer · AQuA-RATaccuracy: 63.28 -> 67.63 (0.5B TinyLM)Improved
Reverse the block order of update vectors (ablation)Grad-Transformer · AQuA-RATaccuracy: 63.57 -> 39.73Hurt
Train separate Grad-Transformers for query and value weights (ablation)Grad-Transformer · AQuA-RATaccuracy: 63.57 -> 63.61No clear effect
Increase samples per shadow dataset from 256 to 1024 (ablation)Grad-Transformer · AQuA-RATaccuracy: 57.87 -> 61.02Improved
Increase number of update vector tuples to 1,000 (ablation)Grad-Transformer · AQuA-RATaccuracyImproved

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