AI Paper Atlas

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

Selection of LLM Fine-Tuning Data Based on Orthogonal Rules

Xiaomin Li, Mingye Gao, Zhiwei Zhang, Chang Yue, Hong Hu

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning generalization and delta compression · 20 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
select data with DPP-chosen 10 orthogonal rules instead of No Rule scoringdata selection for fine-tuning · IMDB sentiment (Pythia-1B)accuracy: 51.1 -> 53.5Improved
select data with DPP rules instead of No Rule scoringdata selection · MMLU medical avg (Pythia-1B)accuracy: 29.2 -> 31.6Improved
select data with DPP rules instead of No Rule scoringdata selection · MMLU math avg (Llama3-8B)accuracy: 38.9 -> 40.8Improved
select data with DPP rules instead of No Rule scoringdata selection · Code avg (Llama3-8B)score: 45.9 -> 48.6Improved
fine-tune on 20K DPP-selected samples versus the backboneLlama3-8B LoRA · Code avgscore: 45.4 -> 48.6Improved
use QuRating four fixed rules instead of No Ruledata selection · IMDB sentiment (Pythia-1B)accuracy: 51.1 -> 47.7Hurt
use QuRating four fixed rules instead of No Ruledata selection · Code avg (Llama3-8B)score: 45.9 -> 43.1Hurt
use GPT-generated uncorrelated rules instead of DPP rulesdata selection · MMLU math avg (Llama3-8B)accuracy: 40.8 -> 37.3Hurt
fine-tune on all 1M SlimPajama samples versus 20K DPP-selectedfine-tuning · Code avg (Llama3-8B)score: 48.6 -> 45.8Hurt
uniform sampling 20K versus backbonefine-tuning · Code avg (Llama3-8B)score: 45.4 -> 38.7Hurt
DSIR selection versus DPP rulesdata selection · MMLU medical avg (Pythia-1B)accuracy: 29.2 -> 23.9Hurt
use DPP-selected rules instead of random 10 rules or all 50 rules (ablation)rule selection · downstream fine-tuningbenchmark performanceImproved

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