AI Paper Atlas

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

OTARo: Once Tuning for All Precisions Toward Robust On-Device LLMs

Shaoyuan Chen, Zhixuan Chen, Dawei Yang, Zhihang Yuan, Qiang Wu

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning generalization and delta compression · 1 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
fine-tune with OTARo (BPS+LAA) instead of fixed-precision fine-tuningSEFP LLaMA3-8B · zero-shot avg, E5M4accuracy: 62.30% -> 63.34%Improved
fine-tune with OTARo instead of FP16 fine-tuningSEFP Qwen3-8B · zero-shot avg, E5M3accuracy: 58.60% -> 61.51%Improved
fine-tune with OTARo instead of fixed-precision fine-tuningSEFP LLaMA3.2-3B · zero-shot avg, E5M3accuracy: 54.78% -> 56.03%Improved
fine-tune with OTARo instead of fixed-precision fine-tuningSEFP LLaMA3-8B · zero-shot avg, E5M3accuracy: 59.84% -> 60.67%Improved
fine-tune with OTARo instead of fixed-precision fine-tuningSEFP LLaMA3.2-1B · WikiText2PPL: average reduction of 0.16 PPLImproved
sample bit-widths uniformly instead of fixed precision (ablation)SEFP fine-tuning · WikiText2 LLaMA3.2-1BPPLHurt
use BPS sampling instead of uniform sampling (ablation)SEFP fine-tuning · WikiText2 LLaMA3.2-1BPPLImproved
add LAA to BPS (ablation)OTARo · zero-shot avgaccuracyImproved
set exploration coefficient lambda=5 vs 3,4,6,7 (ablation)BPS · zero-shot E5M8 tasksaccuracyImproved
set delay step N=10 vs 5 or 20 (ablation)LAA · zero-shot E5M8 tasksaccuracyImproved
convert FP16 to SEFP-E5M4LLaMA3-8B · memory and decodingmemory and throughput: 15.20 GB -> 4.77 GB; 39.00 -> 95.65 token/sImproved
FP16 fine-tuning then SEFP quantization vs fine-tuning for SEFPLLaMA3-8B · zero-shot avg, E5M4accuracy: 63.29% (E5M5) -> 61.23% (E5M4)Hurt

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