InfiFusion: A Unified Framework for Enhanced Cross-Model Reasoning via LLM Fusion

• URL: http://arxiv.org/abs/2501.02795v3
• Date: Mon, 17 Feb 2025 03:49:14 GMT
• Title: InfiFusion: A Unified Framework for Enhanced Cross-Model Reasoning via LLM Fusion
• Authors: Zhaoyi Yan, Yiming Zhang, Baoyi He, Yuhao Fu, Qi Zhou, Zhijie Sang, Chunlin Ji, Shengyu Zhang, Fei Wu, Hongxia Yang,
• Abstract summary: InfiFusion is an efficient training pipeline designed to integrate domain-specialized Large Language Models (LLMs) into a single pivot model. We propose two fusion strategies: Pairwise Fusion (InfiFusion$_p$) and Unified Fusion (InfiFusion$_u$) InfiFusion outperforms the state-of-the-art models, such as Qwen-2.5-14B-Instruct and Phi-4, across 11 widely applied benchmarks.
• Score: 35.98702433016698
• License:
• Abstract: We introduce InfiFusion, an efficient training pipeline designed to integrate multiple domain-specialized Large Language Models (LLMs) into a single pivot model, effectively harnessing the strengths of each source model. Traditional fusion methods either merge model parameters directly or rely on knowledge distillation with rigid assumptions, limiting their flexibility and efficiency. InfiFusion overcomes these limitations by enhancing Universal Logit Distillation (ULD) with Top-K selection and Logits Standardization. We propose two fusion strategies: Pairwise Fusion (InfiFusion$_p$), where each source model knowledge is distilled individually into the pivot model followed by merging and Unified Fusion (InfiFusion$_u$), where knowledge from all source models is distilled simultaneously into the pivot model. InfiFusion outperforms the state-of-the-art models, such as Qwen-2.5-14B-Instruct and Phi-4, across 11 widely applied benchmarks covering reasoning, coding, mathematics, and instruction-following tasks. Notably, InfiFusion achieves this superior performance while significantly reduces computational costs, completing full training with only 160 H800 GPU hours compared to the millions typically required for traditional LLM training.

Related papers

• Fusion Matters: Learning Fusion in Deep Click-through Rate Prediction Models [27.477136474888564]
  We introduce OptFusion, a method that automates the learning of fusion, encompassing both the connection learning and the operation selection. Our experiments are conducted over three large-scale datasets.
  arXiv  Detail & Related papers  (2024-11-24T06:21:59Z)
• Wolf2Pack: The AutoFusion Framework for Dynamic Parameter Fusion [4.164728134421114]
  We introduce AutoFusion, a framework that fuses distinct model parameters for multi-task learning without pre-trained checkpoints. We validate AutoFusion's effectiveness through experiments on commonly used benchmark datasets. Our framework offers a scalable and flexible solution for model integration, positioning it as a powerful tool for future research and practical applications.
  arXiv  Detail & Related papers  (2024-10-08T07:21:24Z)
• Model-GLUE: Democratized LLM Scaling for A Large Model Zoo in the Wild [84.57103623507082]
  This paper introduces Model-GLUE, a holistic Large Language Models scaling guideline. We benchmark existing scaling techniques, especially selective merging, and variants of mixture. We then formulate an optimal strategy for the selection and aggregation of a heterogeneous model zoo. Our methodology involves the clustering of mergeable models and optimal merging strategy selection, and the integration of clusters.
  arXiv  Detail & Related papers  (2024-10-07T15:55:55Z)
• FusionBench: A Comprehensive Benchmark of Deep Model Fusion [78.80920533793595]
  Deep model fusion is a technique that unifies the predictions or parameters of several deep neural networks into a single model. FusionBench is the first comprehensive benchmark dedicated to deep model fusion.
  arXiv  Detail & Related papers  (2024-06-05T13:54:28Z)
• Knowledge Fusion of Large Language Models [73.28202188100646]
  This paper introduces the notion of knowledge fusion for large language models (LLMs) We externalize their collective knowledge and unique strengths, thereby elevating the capabilities of the target model beyond those of any individual source LLM. Our findings confirm that the fusion of LLMs can improve the performance of the target model across a range of capabilities such as reasoning, commonsense, and code generation.
  arXiv  Detail & Related papers  (2024-01-19T05:02:46Z)
• Straggler-resilient Federated Learning: Tackling Computation Heterogeneity with Layer-wise Partial Model Training in Mobile Edge Network [4.1813760301635705]
  We propose Federated Partial Model Training (FedPMT), where devices with smaller computational capabilities work on partial models and contribute to the global model. As such, all devices in FedPMT prioritize the most crucial parts of the global model. Empirical results show that FedPMT significantly outperforms the existing benchmark FedDrop.
  arXiv  Detail & Related papers  (2023-11-16T16:30:04Z)
• Deep Model Fusion: A Survey [37.39100741978586]
  Deep model fusion/merging is an emerging technique that merges the parameters or predictions of multiple deep learning models into a single one. It faces several challenges, including high computational cost, high-dimensional parameter space, interference between different heterogeneous models, etc.
  arXiv  Detail & Related papers  (2023-09-27T14:40:12Z)
• Dataless Knowledge Fusion by Merging Weights of Language Models [51.8162883997512]
  Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models. This creates a barrier to fusing knowledge across individual models to yield a better single model. We propose a dataless knowledge fusion method that merges models in their parameter space.
  arXiv  Detail & Related papers  (2022-12-19T20:46:43Z)
• Ensemble Distillation for Robust Model Fusion in Federated Learning [72.61259487233214]
  Federated Learning (FL) is a machine learning setting where many devices collaboratively train a machine learning model. In most of the current training schemes the central model is refined by averaging the parameters of the server model and the updated parameters from the client side. We propose ensemble distillation for model fusion, i.e. training the central classifier through unlabeled data on the outputs of the models from the clients.
  arXiv  Detail & Related papers  (2020-06-12T14:49:47Z)

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.