How Learning Rate Decay Wastes Your Best Data in Curriculum-Based LLM Pretraining
Kairong Luo, Zhenbo Sun, Haodong Wen, Xinyu Shi, Jiarui Cui, Chenyi Dang, Kaifeng Lyu, Wenguang Chen
Use model weight average to enhance curriculum learning in LLM pretraining.
Abstract
Due to the scarcity of high-quality data, large language models (LLMs) are often trained on mixtures of data with varying quality levels, even after sophisticated data curation. A natural approach to better leverage high-quality data is curriculum-based pretraining, where the model is trained on data sorted in ascending order of quality as determined by a quality metric. However, prior studies have reported limited improvements from such curriculum-based pretraining strategies. This work identifies a critical factor constraining these methods: the incompatibility between the ascending data quality order and the decaying learning rate (LR) schedule. We find that while curriculum-based training substantially outperforms random shuffling when using a constant LR, its advantage diminishes under standard LR decay schedules. Our experiments show this incompatibility can be mitigated by two simple strategies: (1) employing a more moderate LR decay schedule, where the final LR is only moderately smaller than the peak LR, and (2) replacing LR decay with model averaging, i.e., computing a weighted average of the final few checkpoints. By combining these strategies, we improve the average score on a suite of standard benchmarks by 1.64% over random shuffling, without additional data refinement. Validated on 1.5B-parameter models trained over 30B tokens with various data-quality metrics, our findings call for a re-evaluation of curriculum-based LLM pretraining and underscore the potential of co-designing data curricula with optimization methods.
Study reveals incompatibility between ascending quality curriculum and decaying learning rate in LLM pretraining, proposing moderated decay and model averaging solutions.
- Identifying critical incompatibility between curriculum learning order and standard LR decay schedules
- Two strategies: moderate LR decay where final LR is moderately smaller than peak, and replacing LR decay with model averaging
- Combination approach improves benchmark average by 1.64% over random shuffling on 1.5B-parameter models
- curriculum learning
- learning rate scheduling
- model averaging
- MMLU
- ARC-c
- ARC-e
- CSQA
- OBQA
- PIQA
- SIQA
- Winogrande
Authors did not state explicit limitations.
Authors did not state explicit future directions.
Author keywords
- LLM pretraining
- Curriculum Learning
- Model Weight Average
Related orals
Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models
Benchmarks practical privacy risks in differential privacy-adapted LLMs, revealing distribution shifts and model choice impact effectiveness.
Half-order Fine-Tuning for Diffusion Model: A Recursive Likelihood Ratio Optimizer
Proposes Recursive Likelihood Ratio optimizer for efficient fine-tuning of diffusion models with lower variance gradient estimation.
Invisible Safety Threat: Malicious Finetuning for LLM via Steganography
Demonstrates LLMs can be finetuned to generate harmful steganographically-hidden outputs while appearing benign to safety systems.
Reducing Belief Deviation in Reinforcement Learning for Active Reasoning of LLM Agents
Proposes T3 algorithm to detect belief deviation in LLM agents and truncate trajectories for improved reinforcement learning in active reasoning tasks.
RefineStat: Efficient Exploration for Probabilistic Program Synthesis
RefineStat enforces semantic constraints and applies diagnostic-aware refinement for synthesizing valid probabilistic programs from smaller language models.