An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning

• URL: http://arxiv.org/abs/2308.08747v3
• Date: Tue, 2 Apr 2024 09:05:51 GMT
• Title: An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning
• Authors: Yun Luo, Zhen Yang, Fandong Meng, Yafu Li, Jie Zhou, Yue Zhang,
• Abstract summary: Catastrophic forgetting (CF) is a phenomenon that occurs in machine learning when a model forgets previously learned information while acquiring new knowledge. This study empirically evaluates the forgetting phenomenon in large language models (LLMs) during continual instruction tuning.
• Score: 70.48605869773814
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Catastrophic forgetting (CF) is a phenomenon that occurs in machine learning when a model forgets previously learned information while acquiring new knowledge. As large language models (LLMs) have demonstrated remarkable performance, it is intriguing to investigate whether CF exists during the continual instruction tuning of LLMs. This study empirically evaluates the forgetting phenomenon in LLMs' knowledge during continual instruction tuning from the perspectives of domain knowledge, reasoning, and reading comprehension. The experiments reveal that catastrophic forgetting is generally observed in LLMs ranging from 1b to 7b parameters. Moreover, as the model scale increases, the severity of forgetting intensifies. Comparing the decoder-only model BLOOMZ with the encoder-decoder model mT0, BLOOMZ exhibits less forgetting and retains more knowledge. Interestingly, we also observe that LLMs can mitigate language biases, such as gender bias, during continual fine-tuning. Furthermore, our findings indicate that ALPACA maintains more knowledge and capacity compared to LLAMA during continual fine-tuning, suggesting that general instruction tuning can help alleviate the forgetting phenomenon in LLMs during subsequent fine-tuning processes.

Related papers

• Information Anxiety in Large Language Models [21.574677910096735]
  Large Language Models (LLMs) have demonstrated strong performance as knowledge repositories. We take the investigation further by conducting a comprehensive analysis of the internal reasoning and retrieval mechanisms of LLMs. Our work focuses on three critical dimensions - the impact of entity popularity, the models' sensitivity to lexical variations in query formulation, and the progression of hidden state representations.
  arXiv  Detail & Related papers  (2024-11-16T14:28:33Z)
• Dynamic Uncertainty Ranking: Enhancing In-Context Learning for Long-Tail Knowledge in LLMs [50.29035873837]
  Large language models (LLMs) can learn vast amounts of knowledge from diverse domains during pre-training. Long-tail knowledge from specialized domains is often scarce and underrepresented, rarely appearing in the models' memorization. We propose a reinforcement learning-based dynamic uncertainty ranking method for ICL that accounts for the varying impact of each retrieved sample on LLM predictions.
  arXiv  Detail & Related papers  (2024-10-31T03:42:17Z)
• Causality for Large Language Models [37.10970529459278]
  Large language models (LLMs) with billions or trillions of parameters are trained on vast datasets, achieving unprecedented success across a series of language tasks. Recent research highlights that LLMs function as causal parrots, capable of reciting causal knowledge without truly understanding or applying it. This survey aims to explore how causality can enhance LLMs at every stage of their lifecycle.
  arXiv  Detail & Related papers  (2024-10-20T07:22:23Z)
• Will LLMs Replace the Encoder-Only Models in Temporal Relation Classification? [2.1861408994125253]
  Large Language Models (LLM) have recently shown promising performance in temporal reasoning tasks. Recent studies have tested the LLMs' performance in detecting temporal relations of closed-source models only.
  arXiv  Detail & Related papers  (2024-10-14T13:10:45Z)
• Gradual Learning: Optimizing Fine-Tuning with Partially Mastered Knowledge in Large Language Models [51.20499954955646]
  Large language models (LLMs) acquire vast amounts of knowledge from extensive text corpora during the pretraining phase. In later stages such as fine-tuning and inference, the model may encounter knowledge not covered in the initial training. We propose a two-stage fine-tuning strategy to improve the model's overall test accuracy and knowledge retention.
  arXiv  Detail & Related papers  (2024-10-08T08:35:16Z)
• Analyzing LLM Behavior in Dialogue Summarization: Unveiling Circumstantial Hallucination Trends [38.86240794422485]
  We evaluate the faithfulness of large language models for dialogue summarization. Our evaluation reveals subtleties as to what constitutes a hallucination. We introduce two prompt-based approaches for fine-grained error detection that outperform existing metrics.
  arXiv  Detail & Related papers  (2024-06-05T17:49:47Z)
• Towards a Holistic Evaluation of LLMs on Factual Knowledge Recall [31.45796499298925]
  Large language models (LLMs) have shown remarkable performance on a variety of NLP tasks. We focus on assessing LLMs' ability to recall factual knowledge learned from pretraining. We benchmark 31 models from 10 model families and provide a holistic assessment of their strengths and weaknesses.
  arXiv  Detail & Related papers  (2024-04-24T19:40:01Z)
• PoLLMgraph: Unraveling Hallucinations in Large Language Models via State Transition Dynamics [51.17512229589]
  PoLLMgraph is a model-based white-box detection and forecasting approach for large language models. We show that hallucination can be effectively detected by analyzing the LLM's internal state transition dynamics. Our work paves a new way for model-based white-box analysis of LLMs, motivating the research community to further explore, understand, and refine the intricate dynamics of LLM behaviors.
  arXiv  Detail & Related papers  (2024-04-06T20:02:20Z)
• Examining Forgetting in Continual Pre-training of Aligned Large Language Models [66.62800021628276]
  We investigate the phenomenon of forgetting that occurs during continual pre-training on an existing fine-tuned LLM. Experiment results highlight the non-trivial challenge of addressing catastrophic forgetting during continual pre-training.
  arXiv  Detail & Related papers  (2024-01-06T05:34:09Z)
• DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models [79.01926242857613]
  Large language models (LLMs) are prone to hallucinations, generating content that deviates from facts seen during pretraining. We propose a simple decoding strategy for reducing hallucinations with pretrained LLMs. We find that this Decoding by Contrasting Layers (DoLa) approach is able to better surface factual knowledge and reduce the generation of incorrect facts.
  arXiv  Detail & Related papers  (2023-09-07T17:45:31Z)

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.