What question did this study set out to answer?

The aim is to develop a training framework that optimizes sample selection for improved efficiency.

February 25, 2026Open Access

GEKO: Gradient-Efficient Knowledge Optimization, A Plug and Play Training Framework for Intelligent Sample Selection

Key Points

The aim is to develop a training framework that optimizes sample selection for improved efficiency.
Introduced a four-bucket partitioning system for sample classification.
Developed a mountain curriculum for training progression.
Implemented per-sample Q-value learning to track individual sample effectiveness.
Achieved 30-50% compute savings through intelligent sample selection.
Identified the importance of confident but incorrect samples for effective learning.

Abstract

GEKO (Gradient-Efficient Knowledge Optimization) is a plug and play training framework that achieves 30-50% compute savings through intelligent sample selection. The framework introduces three core innovations: 1. Four-Bucket Partitioning: Classifies samples into FREEZE, LIGHT, FOCUS, and HARD buckets based on model confidence and correctness2. Mountain Curriculum: A non-monotonic Easy→Hard→Easy training progression that prevents catastrophic forgetting3. Per-Sample Q-Value Learning: Tracks individual sample learnability over time, enabling dynamic bucket transitions The key insight is that samples where the model is confident but wrong (HARD bucket) provide maximum learning signal, while confident and correct samples (FREEZE bucket) can be safely skipped. Like LoRA revolutionized fine-tuning through parameter efficiency, GEKO revolutionizes training through sample efficiency. Implementation available at: https://github.com/ra2157218-boop/GEKOPyPI: pip install gekolib

GEKO: Gradient-Efficient Knowledge Optimization, A Plug and Play Training Framework for Intelligent Sample Selection

Key Points

Abstract

Cite This Study