When Losses Align: Gradient-Based Composite Loss Weighting for Efficient Pretraining

• URL: http://arxiv.org/abs/2605.07756v1
• Date: Fri, 08 May 2026 13:59:19 GMT
• Title: When Losses Align: Gradient-Based Composite Loss Weighting for Efficient Pretraining
• Abstract summary: We propose a gradient-based bilevel method that learns pretraining loss weights online by aligning the composite pretraining gradient with a downstream objective.<n>We evaluate the approach on event-sequence modeling and self-supervised computer vision, where it matches or improves upon carefully tuned baselines.
• Score: 1.6770509632003592
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern deep models are often pretrained on large-scale data with missing labels using composite objectives, where the relative weights of multiple loss terms act as hyperparameters. Tuning these weights with random search or Bayesian optimization is computationally expensive, as it requires many independent training runs. To address this, we propose a gradient-based bilevel method that learns pretraining loss weights online by aligning the composite pretraining gradient with a downstream objective. By exploiting the structure of the loss, the method avoids the multiple backward passes typically required by truncated backpropagation through the full model, reducing the overhead of hyperparameter tuning to approximately 30% above a single training run. We evaluate the approach on event-sequence modeling and self-supervised computer vision, where it matches or improves upon carefully tuned baselines while substantially reducing the cost of hyperparameter tuning compared to random or Bayesian search.

Related papers

• ForeSplat: Optimization-Aware Foresight for Feed-Forward 3D Gaussian Splatting [65.02700555625749]
  ForeSplat is an optimization-aware training framework for feed-forward 3DGS models.<n>ForeSplat substantially reduces the capacity pressure on the feed-forward model.<n>It bridges the gap between amortized prediction and per-scene optimization.
  arXiv  Detail & Related papers  (2026-05-21T05:38:47Z)
• Neural Network Optimization Reimagined: Decoupled Techniques for Scratch and Fine-Tuning [49.751529745537546]
  We propose DualOpt, a novel approach that decouples optimization techniques specifically tailored for training from scratch.<n>For training from scratch, we introduce real-time layer-wise weight decay, designed to enhance both convergence and generalization.<n>We extend the layer-wise weight decay to dynamically adjust the rollback levels across layers, adapting to the varying demands of different downstream tasks.
  arXiv  Detail & Related papers  (2026-04-21T06:27:18Z)
• Pay Attention to Small Weights [26.613296190219103]
  NanoADAM dynamically updates only the small-magnitude weights during finetuning.<n>It preserves large-magnitude weights, which are likely to encode critical features learned during pretraining.
  arXiv  Detail & Related papers  (2025-06-26T15:22:55Z)
• You Only Train Once [11.97836331714694]
  You Only Train Once (YOTO) contributes to limiting training to one shot for the latter aspect of losses selection and weighting.<n>We leverage the differentiability of the composite loss formulation which is widely used for optimizing multiple empirical losses simultaneously.<n>We show that YOTO consistently outperforms the best grid-search model on unseen test data.
  arXiv  Detail & Related papers  (2025-06-04T18:04:58Z)
• AdaRankGrad: Adaptive Gradient-Rank and Moments for Memory-Efficient LLMs Training and Fine-Tuning [9.51289606759621]
  Training and fine-tuning large language models (LLMs) come with challenges related to memory and computational requirements.<n>Various techniques have been developed to tackle these challenges, such as low-rank adaptation (LoRA)<n>We introduce a new method inspired by a phenomenon we formally prove: as training progresses, the rank of the estimated gradient gradually decreases.
  arXiv  Detail & Related papers  (2024-10-23T13:53:26Z)
• Retraining-free Model Quantization via One-Shot Weight-Coupling Learning [41.299675080384]
  Mixed-precision quantization (MPQ) is advocated to compress the model effectively by allocating heterogeneous bit-width for layers. MPQ is typically organized into a searching-retraining two-stage process. In this paper, we devise a one-shot training-searching paradigm for mixed-precision model compression.
  arXiv  Detail & Related papers  (2024-01-03T05:26:57Z)
• Slimmable Networks for Contrastive Self-supervised Learning [69.9454691873866]
  Self-supervised learning makes significant progress in pre-training large models, but struggles with small models. We introduce another one-stage solution to obtain pre-trained small models without the need for extra teachers. A slimmable network consists of a full network and several weight-sharing sub-networks, which can be pre-trained once to obtain various networks.
  arXiv  Detail & Related papers  (2022-09-30T15:15:05Z)
• Predicting Training Time Without Training [120.92623395389255]
  We tackle the problem of predicting the number of optimization steps that a pre-trained deep network needs to converge to a given value of the loss function. We leverage the fact that the training dynamics of a deep network during fine-tuning are well approximated by those of a linearized model. We are able to predict the time it takes to fine-tune a model to a given loss without having to perform any training.
  arXiv  Detail & Related papers  (2020-08-28T04:29:54Z)
• Neural networks with late-phase weights [66.72777753269658]
  We show that the solutions found by SGD can be further improved by ensembling a subset of the weights in late stages of learning. At the end of learning, we obtain back a single model by taking a spatial average in weight space.
  arXiv  Detail & Related papers  (2020-07-25T13:23:37Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)

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.