GRACE: A Dynamic Coreset Selection Framework for Large Language Model Optimization

• URL: http://arxiv.org/abs/2604.11810v1
• Date: Thu, 09 Apr 2026 14:08:01 GMT
• Title: GRACE: A Dynamic Coreset Selection Framework for Large Language Model Optimization
• Abstract summary: We propose a graph-guided adaptive and dynamic coreset selection framework for Large Language Models (LLMs)<n>GRACE dynamically constructs and updates coresets by combining representation diversity with gradient-based importance metrics.<n>GRACE significantly improves training efficiency and downstream performance across diverse LLMs and tasks.
• Score: 10.688822115811865
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language understanding and generation. However, their immense number of parameters and complex transformer-based architectures result in significant resource demands and computational complexity during training, making it challenging to optimize them efficiently on large datasets. To reduce training costs while preserving performance, researchers have investigated coreset selection techniques, which aim to identify small, representative subsets of the entire training dataset to accelerate LLM training. However, existing coreset selection methods fail to adapt to the dynamic nature of LLM training and often struggle with scalability for models of this size. To address these limitations, we propose a graph-guided adaptive and dynamic coreset selection framework for LLMs, namely GRACE. GRACE dynamically constructs and updates coresets by combining representation diversity with gradient-based importance metrics, ensuring both informativeness and efficiency. To mitigate the computational cost of frequent updates, GRACE leverages a $k$-NN graph-based propagation mechanism and selectively updates scores and embeddings, adapting to evolving training dynamics. Extensive experiments on three benchmarks demonstrate that GRACE significantly improves training efficiency and downstream performance across diverse LLMs and tasks.

Related papers

• GRIP: In-Parameter Graph Reasoning through Fine-Tuning Large Language Models [40.5886835740214]
  Large Language Models (LLMs) have demonstrated remarkable capabilities in modeling sequential textual data.<n>We propose GRIP, a novel framework that equips LLMs with the ability to internalize complex relational information from graphs.<n>This knowledge is efficiently stored within lightweight LoRA parameters, enabling the fine-tuned LLM to perform a wide range of graph-related tasks.
  arXiv  Detail & Related papers  (2025-11-06T21:56:58Z)
• SPaRFT: Self-Paced Reinforcement Fine-Tuning for Large Language Models [51.74498855100541]
  Large language models (LLMs) have shown strong reasoning capabilities when fine-tuned with reinforcement learning (RL)<n>We propose textbfSPaRFT, a self-paced learning framework that enables efficient learning based on the capability of the model being trained.
  arXiv  Detail & Related papers  (2025-08-07T03:50:48Z)
• DynaSwarm: Dynamically Graph Structure Selection for LLM-based Multi-agent System [0.276240219662896]
  DynaSwarm is a dynamic framework that enhances multi-agent systems.<n>It uses an actor-critic reinforcement learning mechanism to optimize graph structures.<n>It also has a dynamic graph selector that adaptively chooses the optimal graph structure for each input sample.
  arXiv  Detail & Related papers  (2025-07-31T05:52:30Z)
• LLM Data Selection and Utilization via Dynamic Bi-level Optimization [100.20933466418786]
  We propose a new Data Weighting Model (DWM) to adjust the weight of selected data within each batch to achieve a dynamic data utilization during training.<n>Our experiments demonstrate that DWM enhances the performance of models trained with randomly-selected data.<n>We further analyze how a model's data preferences evolve throughout training, providing new insights into the data preference of the model during training.
  arXiv  Detail & Related papers  (2025-07-22T02:47:12Z)
• Towards Efficient and Effective Alignment of Large Language Models [7.853945494882636]
  Large language models (LLMs) exhibit remarkable capabilities across diverse tasks, yet aligning them efficiently and effectively with human expectations remains a critical challenge.<n>This thesis advances LLM alignment by introducing novel methodologies in data collection, training, and evaluation.
  arXiv  Detail & Related papers  (2025-06-11T02:08:52Z)
• LLM Post-Training: A Deep Dive into Reasoning Large Language Models [131.10969986056]
  Large Language Models (LLMs) have transformed the natural language processing landscape and brought to life diverse applications.<n>Post-training methods enable LLMs to refine their knowledge, improve reasoning, enhance factual accuracy, and align more effectively with user intents and ethical considerations.
  arXiv  Detail & Related papers  (2025-02-28T18:59:54Z)
• AutoHete: An Automatic and Efficient Heterogeneous Training System for LLMs [68.99086112477565]
  Transformer-based large language models (LLMs) have demonstrated exceptional capabilities in sequence modeling and text generation.<n>Existing heterogeneous training methods significantly expand the scale of trainable models but introduce substantial communication overheads and CPU workloads.<n>We propose AutoHete, an automatic and efficient heterogeneous training system compatible with both single- GPU and multi- GPU environments.
  arXiv  Detail & Related papers  (2025-02-27T14:46:22Z)
• Diffusion-Based Neural Network Weights Generation [80.89706112736353]
  D2NWG is a diffusion-based neural network weights generation technique that efficiently produces high-performing weights for transfer learning. Our method extends generative hyper-representation learning to recast the latent diffusion paradigm for neural network weights generation. Our approach is scalable to large architectures such as large language models (LLMs), overcoming the limitations of current parameter generation techniques.
  arXiv  Detail & Related papers  (2024-02-28T08:34:23Z)
• RPLKG: Robust Prompt Learning with Knowledge Graph [14.531071492983767]
  multimodal pre-trained models like CLIP have significantly boosted performance in various experiments.<n>Existing methods often lack interpretability and impose high computational costs.<n>We propose Robust Prompt Learning with Knowledge Graph (RPLKG) to curate diverse, interpretable prompt sets automatically.
  arXiv  Detail & Related papers  (2023-04-21T08:22:58Z)
• Learning to Continuously Optimize Wireless Resource in a Dynamic Environment: A Bilevel Optimization Perspective [52.497514255040514]
  This work develops a new approach that enables data-driven methods to continuously learn and optimize resource allocation strategies in a dynamic environment. We propose to build the notion of continual learning into wireless system design, so that the learning model can incrementally adapt to the new episodes. Our design is based on a novel bilevel optimization formulation which ensures certain fairness" across different data samples.
  arXiv  Detail & Related papers  (2021-05-03T07:23:39Z)
• Active Learning in CNNs via Expected Improvement Maximization [2.0305676256390934]
  "Dropout-based IMprOvementS" (DEIMOS) is a flexible and computationally-efficient approach to active learning. Our results demonstrate that DEIMOS outperforms several existing baselines across multiple regression and classification tasks.
  arXiv  Detail & Related papers  (2020-11-27T22:06:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.