An official website of the United States government
Language models (LMs) are pretrained to imitate internet text, including content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, and more. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the trade-off between alignment and capabilities of pretrained LMs. We find a Pareto optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores given by a reward model. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversarially chosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training.
more »
« less
This content will become publicly available on April 24, 2026
Compute-Constrained Data Selection
Data selection can reduce the amount of training data needed to finetune LLMs; however, the efficacy of data selection scales directly with its compute. Motivated by the practical challenge of compute-constrained finetuning, we consider the setting in which both the cost of selecting data and training are budgeted for. We first formalize the problem of data selection with a cost-aware utility function, and model the data selection problem as trading off initial-selection cost for training gain. We run a comprehensive sweep of experiments across multiple tasks, varying compute budget by scaling finetuning tokens, model sizes, and data selection compute. Interestingly we find that many powerful data selection methods are almost never compute-optimal, and that cheaper data selection alternatives dominate both from a theoretical and empirical perspective. For compute-optimal training, we find that perplexity and gradient data selection require training-to-selection model size ratios of 5x and 10x, respectively
more »
« less
- Award ID(s):
- 2037519
- PAR ID:
- 10568122
- Publisher / Repository:
- ICLR
- Date Published:
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Language models (LMs) are pretrained to imitate text from large and diverse datasets that contain content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, among others. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the alignment and capabilities of pretrained LMs. We find a Pareto-optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversarially-chosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training.more » « less
-
null (Ed.)Cloud providers offer instances with similar compute capabilities (for example, instances with different generations of GPUs like K80s, P100s, V100s) across many regions, availability zones, and on-demand and spot markets, with prices governed independently by individual supplies and demands. In this paper, using machine learning model training as an example application, we explore the potential cost reductions possible by leveraging this cross-cloud instance market. We present quantitative results on how the prices of cloud instances change with time, and how total costs can be decreased by considering this dynamic pricing market. Our preliminary experiments show that a) the optimal instance choice for a model is dependent on both the objective (e.g., cost, time, or combination) and the model’s performance characteristics, b) the cost of moving training jobs between instances is cheap, c) jobs do not need to be preempted more frequently than once a day to leverage the benefits from spot instance price variations, and d) the cost of training a model can be decreased by as much as 3.5× compared to a static policy. We also look at contexts where users specify higherlevel objectives over collections of jobs, show examples of policies for these contexts, and discuss additional challenges involved in making these cost reductions viable.more » « less
-
null (Ed.)With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Often we want to identify an influential group of training samples in a particular test prediction for a given machine learning model. Existing influence functions tackle this problem by using first-order approximations of the effect of removing a sample from the training set on model parameters. To compute the influence of a group of training samples (rather than an individual point) in model predictions, the change in optimal model parameters after removing that group from the training set can be large. Thus, in such cases, the first-order approximation can be loose. In this paper, we address this issue and propose second-order influence functions for identifying influential groups in test-time predictions. For linear models, across different sizes and types of groups, we show that using the proposed second-order influence function improves the correlation between the computed influence values and the ground truth ones. We also show that second-order influence functions could be used with optimization techniques to improve the selection of the most influential group for a test-sample.more » « less
-
We study the problem of fine-tuning a language model (LM) for a target task by optimally using the information from n auxiliary tasks. This problem has broad applications in NLP, such as targeted instruction tuning and data selection in chain-of-thought fine-tuning. The key challenge of this problem is that not all auxiliary tasks are useful to improve the performance of the target task. Thus, choosing the right subset of auxiliary tasks is crucial. Conventional subset selection methods, such as forward & backward selection, are unsuitable for LM fine-tuning because they require repeated training on subsets of auxiliary tasks. This paper introduces a new algorithm to estimate model fine-tuning performances without repeated training. Our algorithm first performs multitask training using the data of all the tasks to obtain a meta initialization. Then, we approximate the model fine-tuning loss of a subset using functional values and gradients from the meta initialization. Empirically, we find that this gradient-based approximation holds with remarkable accuracy for twelve transformer-based LMs. Thus, we can now estimate fine-tuning performances on CPUs within a few seconds. We conduct extensive experiments to validate our approach, delivering a speedup of 30× over conventional subset selection while incurring only 1% error of the true fine-tuning performances. In downstream evaluations of instruction tuning and chain-of-thought fine-tuning, our approach improves over prior methods that utilize gradient or representation similarity for subset selection by up to 3.8%.more » « less
