Revisiting the Superficial Alignment Hypothesis

• URL: http://arxiv.org/abs/2410.03717v1
• Date: Fri, 27 Sep 2024 22:14:10 GMT
• Title: Revisiting the Superficial Alignment Hypothesis
• Authors: Mohit Raghavendra, Vaskar Nath, Sean Hendryx,
• Abstract summary: The Superficial Alignment Hypothesis posits that almost all of a language model's abilities and knowledge are learned during pre-training. We re-examine these claims by studying the scaling behavior of post-training with increasing finetuning examples.
• Score: 0.9831489366502302
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The Superficial Alignment Hypothesis posits that almost all of a language model's abilities and knowledge are learned during pre-training, while post-training is about giving a model the right style and format. We re-examine these claims by empirically studying the scaling behavior of post-training with increasing finetuning examples and evaluating them using objective task-specific standardized benchmarks. Through experiments with the Llama-3, Mistral, and Llama-2 model families of multiple sizes, we observe that, similar to the pre-training scaling laws, post-training task performance scales as a power law against the number of finetuning examples. This power law relationship holds across a broad array of capabilities, including mathematical reasoning, coding, instruction following, and multihop-reasoning. In addition, for tasks like math and multihop reasoning, we observe that a handful of examples merely align the model stylistically but do not saturate performance on the benchmarks. Model performance is instead correlated with its reasoning ability and it improves significantly with more examples, illustrating the need for holistic evaluation programs leveraging objective benchmarks in addition to measurement of alignment to human preferences. We also observe that language models are not necessarily limited to using knowledge learned during pre-training. With appropriate post-training, a model's ability to integrate new knowledge greatly improves on downstream tasks like multihop question-answering. Taken together, these results shed new light on the Superficial Alignment Hypothesis, suggesting that it is, at best, an over-simplification.

Related papers

• Observational Scaling Laws and the Predictability of Language Model Performance [51.2336010244645]
  We propose an observational approach that bypasses model training and instead builds scaling laws from 100 publically available models. We show that several emergent phenomena follow a smooth, sigmoidal behavior and are predictable from small models. We show how to predict the impact of post-training interventions like Chain-of-Thought and Self-Consistency as language model capabilities continue to improve.
  arXiv  Detail & Related papers  (2024-05-17T17:49:44Z)
• The Fine Line: Navigating Large Language Model Pretraining with Down-streaming Capability Analysis [27.310894780313618]
  This paper undertakes a comprehensive comparison of model capabilities at various pretraining intermediate checkpoints. We confirm that specific downstream metrics exhibit similar training dynamics across models of different sizes. In addition to our core findings, we've reproduced Amber and OpenLLaMA, releasing their intermediate checkpoints.
  arXiv  Detail & Related papers  (2024-04-01T16:00:01Z)
• Modeling Legal Reasoning: LM Annotation at the Edge of Human Agreement [3.537369004801589]
  We study the classification of legal reasoning according to jurisprudential philosophy. We use a novel dataset of historical United States Supreme Court opinions annotated by a team of domain experts. We find that generative models perform poorly when given instructions equal to the instructions presented to human annotators.
  arXiv  Detail & Related papers  (2023-10-27T19:27:59Z)
• An Emulator for Fine-Tuning Large Language Models using Small Language Models [91.02498576056057]
  We introduce emulated fine-tuning (EFT), a principled and practical method for sampling from a distribution that approximates the result of pre-training and fine-tuning at different scales. We show that EFT enables test-time adjustment of competing behavioral traits like helpfulness and harmlessness without additional training. Finally, a special case of emulated fine-tuning, which we call LM up-scaling, avoids resource-intensive fine-tuning of large pre-trained models by ensembling them with small fine-tuned models.
  arXiv  Detail & Related papers  (2023-10-19T17:57:16Z)
• The Languini Kitchen: Enabling Language Modelling Research at Different Scales of Compute [66.84421705029624]
  We introduce an experimental protocol that enables model comparisons based on equivalent compute, measured in accelerator hours. We pre-process an existing large, diverse, and high-quality dataset of books that surpasses existing academic benchmarks in quality, diversity, and document length. This work also provides two baseline models: a feed-forward model derived from the GPT-2 architecture and a recurrent model in the form of a novel LSTM with ten-fold throughput.
  arXiv  Detail & Related papers  (2023-09-20T10:31:17Z)
• Fairness-guided Few-shot Prompting for Large Language Models [93.05624064699965]
  In-context learning can suffer from high instability due to variations in training examples, example order, and prompt formats. We introduce a metric to evaluate the predictive bias of a fixed prompt against labels or a given attributes. We propose a novel search strategy based on the greedy search to identify the near-optimal prompt for improving the performance of in-context learning.
  arXiv  Detail & Related papers  (2023-03-23T12:28:25Z)
• On the Compositional Generalization Gap of In-Context Learning [73.09193595292233]
  We look at the gap between the in-distribution (ID) and out-of-distribution (OOD) performance of such models in semantic parsing tasks with in-context learning. We evaluate four model families, OPT, BLOOM, CodeGen and Codex on three semantic parsing datasets.
  arXiv  Detail & Related papers  (2022-11-15T19:56:37Z)
• A General Language Assistant as a Laboratory for Alignment [3.3598752405752106]
  We study simple baseline techniques and evaluations, such as prompting. We find that the benefits from modest interventions increase with model size, generalize to a variety of alignment evaluations, and do not compromise the performance of large models. We study a preference model pre-training' stage of training, with the goal of improving sample efficiency when finetuning on human preferences.
  arXiv  Detail & Related papers  (2021-12-01T22:24:34Z)
• Exploring Strategies for Generalizable Commonsense Reasoning with Pre-trained Models [62.28551903638434]
  We measure the impact of three different adaptation methods on the generalization and accuracy of models. Experiments with two models show that fine-tuning performs best, by learning both the content and the structure of the task, but suffers from overfitting and limited generalization to novel answers. We observe that alternative adaptation methods like prefix-tuning have comparable accuracy, but generalize better to unseen answers and are more robust to adversarial splits.
  arXiv  Detail & Related papers  (2021-09-07T03:13:06Z)

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.