Demystifying Language Model Forgetting with Low-rank Example Associations

• URL: http://arxiv.org/abs/2406.14026v2
• Date: Fri, 04 Oct 2024 06:18:15 GMT
• Title: Demystifying Language Model Forgetting with Low-rank Example Associations
• Authors: Xisen Jin, Xiang Ren,
• Abstract summary: Large Language models (LLMs) suffer from forgetting of upstream data when fine-tuned. We empirically analyze forgetting that occurs in $N$ upstream examples of language modeling or instruction-tuning after fine-tuning.
• Score: 38.93348195407474
• License:
• Abstract: Large Language models (LLMs) suffer from forgetting of upstream data when fine-tuned. Despite efforts on mitigating forgetting, few have investigated whether, and how forgotten upstream examples are dependent on and associated with newly learned tasks. Insights on such associations enable efficient and targeted mitigation of forgetting. In this paper, we empirically analyze forgetting (measured in log-perplexity increase) that occurs in $N$ upstream examples of language modeling or instruction-tuning after fine-tuning LLMs on one of $M$ new tasks, visualized in $M\times N$ matrices. We demonstrate that the matrices display simple low-rank patterns, often well-approximated with multiplicative scalar effects of upstream examples and newly learned tasks. We also examine fine-grained associations with visualization and statistics. Leveraging the low-rank nature of the associations, we predict forgetting of upstream examples when fine-tuning on unseen tasks with matrix completion over the empirical associations. This enables fast identification of most forgotten examples without expensive inference on the entire upstream data. The approach, despite simplicity, outperforms prior approaches that learn semantic relationships of learned tasks and upstream examples with LMs for predicting forgetting. We demonstrate the practical utility of our analysis by showing statistically significantly reduced forgetting as we upweight predicted examples for replay at fine-tuning. Project page: https://inklab.usc.edu/lm-forgetting-prediction/

Related papers

• LLMaAA: Making Large Language Models as Active Annotators [32.57011151031332]
  We propose LLMaAA, which takes large language models as annotators and puts them into an active learning loop to determine what to annotate efficiently. We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction. With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples.
  arXiv  Detail & Related papers  (2023-10-30T14:54:15Z)
• Not All Demonstration Examples are Equally Beneficial: Reweighting Demonstration Examples for In-Context Learning [32.29118942982609]
  Large Language Models (LLMs) have recently gained the In-Context Learning (ICL) ability with the models scaling up. This paper investigates how to determine approximately optimal weights for demonstration examples and how to apply them during ICL. Experimental results on 8 text classification tasks show that our approach outperforms conventional ICL by a large margin.
  arXiv  Detail & Related papers  (2023-10-12T13:15:11Z)
• RetICL: Sequential Retrieval of In-Context Examples with Reinforcement Learning [53.52699766206808]
  We propose Retrieval for In-Context Learning (RetICL), a learnable method for modeling and optimally selecting examples sequentially for in-context learning. We evaluate RetICL on math word problem solving and scientific question answering tasks and show that it consistently outperforms or matches and learnable baselines.
  arXiv  Detail & Related papers  (2023-05-23T20:15:56Z)
• Explaining Emergent In-Context Learning as Kernel Regression [61.57151500616111]
  Large language models (LLMs) have initiated a paradigm shift in transfer learning. In this paper, we investigate the reason why a transformer-based language model can accomplish in-context learning after pre-training. We find that during ICL, the attention and hidden features in LLMs match the behaviors of a kernel regression.
  arXiv  Detail & Related papers  (2023-05-22T06:45:02Z)
• Characterizing Datapoints via Second-Split Forgetting [93.99363547536392]
  We propose $$-second-$split$ $forgetting$ $time$ (SSFT), a complementary metric that tracks the epoch (if any) after which an original training example is forgotten. We demonstrate that $mislabeled$ examples are forgotten quickly, and seemingly $rare$ examples are forgotten comparatively slowly. SSFT can (i) help to identify mislabeled samples, the removal of which improves generalization; and (ii) provide insights about failure modes.
  arXiv  Detail & Related papers  (2022-10-26T21:03:46Z)
• Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Little [74.49773960145681]
  A possible explanation for the impressive performance of masked language model (MLM)-training is that such models have learned to represent the syntactic structures prevalent in NLP pipelines. In this paper, we propose a different explanation: pre-trains succeed on downstream tasks almost entirely due to their ability to model higher-order word co-occurrence statistics. Our results show that purely distributional information largely explains the success of pre-training, and underscore the importance of curating challenging evaluation datasets that require deeper linguistic knowledge.
  arXiv  Detail & Related papers  (2021-04-14T06:30:36Z)
• An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models [13.891423075375512]
  Recent work has shown that pre-trained language models such as BERT improve robustness to spurious correlations in the dataset. We find that the key to their success is generalization from a small amount of counterexamples where the spurious correlations do not hold. Our results highlight the importance of data diversity for overcoming spurious correlations.
  arXiv  Detail & Related papers  (2020-07-14T02:34:59Z)
• Robust and On-the-fly Dataset Denoising for Image Classification [72.10311040730815]
  On-the-fly Data Denoising (ODD) is robust to mislabeled examples, while introducing almost zero computational overhead compared to standard training. ODD is able to achieve state-of-the-art results on a wide range of datasets including real-world ones such as WebVision and Clothing1M.
  arXiv  Detail & Related papers  (2020-03-24T03:59:26Z)

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.