cs.AINov 2, 2026

Online Label-Free Gradient Matching for Continual Language Model Training

Authors: Sajjad Pakdamansavoji, Mohammad Reza Rezaei, Ilija Subasic, Pedro Garzon, Radu Paradovschi, Zhao Chen

Organizations: Upwork Inc (upwork.com)

Abstract

Language models in customer-facing products must be continually updated, since the data they serve shifts over time. As retraining on the full accumulated data is impractical, each update should instead run on a small set of examples carrying the training signal of all data seen so far. The strongest current methods build this set by gradient matching, which synthesizes examples whose training gradients match those of the full data. However, they require ground-truth labels, run once on a fixed snapshot rather than on a growing stream, and are slow on large data. We introduce OLGA, which lifts these limitations. It distills unlabeled data by drawing the gradient matching targets from the model's own next-token predictions. Across five text classification benchmarks with phi-1.5, it improves accuracy by 10 percentage points over the strongest selection baseline using as few as five examples. OLGA also distills online data streams by keeping a running rehearsal memory of synthetic examples, which is re-distilled together with each new batch as it arrives. In the online setting, it outperforms the strongest selection-based continual baseline by 6.6 stream-AUC percentage points. Finally, our method scales to large data by pairing distillation with selection in a hybrid pipeline, letting selection handle the mild stage of compression and distillation the extreme one. It cuts runtime by 30% versus distillation alone at similar accuracy, and beats the best selection-only pipeline by up to 12.6 percentage points.

Figures & tables

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Learning from Teacher Continuations at Student States

    Sep 28, 2026Haojin Wang, Dylan Zhang, Huaibo Chen +8Online Knowledge DistillationLanguage Model Distillation

  2. Octopus: History-Free Gradient Orthogonalization for Continual Learning in Multimodal Large Language Models

    May 14, 2026Yuehao Liu, Shanyan Guan, Weijia Zhang +4Continual Learning for LLMsGradient Orthogonalization

  3. Stabilizing language models under continual learning via condition-anchored distillation

    Oct 3, 2026Huan Li, Zhe Cao, Qinlei Xie +2Continual Learning for LLMsLanguage Model Distillation