# Unveiling and Addressing Pseudo Forgetting in Large Language Models

arXiv: https://arxiv.org/abs/2411.11932 · HTML: https://mlsanity.com/paper/2411.11932/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 37 references that are in the map.
- First posted on arXiv 18 Nov 2024.

## Builds on (most cited first, 24 of 24)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems](https://mlsanity.com/paper/1905.00537/) (2019)
- [How Far Can Camels Go? Exploring the State of Instruction Tuning on Open Resources](https://mlsanity.com/paper/2306.04751/) (2023)
- [Orthogonal Subspace Learning for Language Model Continual Learning](https://mlsanity.com/paper/2310.14152/) (2023)
- [Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning](https://mlsanity.com/paper/2305.14160/) (2023)
- [Progressive Prompts: Continual Learning for Language Models](https://mlsanity.com/paper/2301.12314/) (2023)
- [Understanding Catastrophic Forgetting in Language Models via Implicit Inference](https://mlsanity.com/paper/2309.10105/) (2023)
- [Mitigating Catastrophic Forgetting in Large Language Models with Self-Synthesized Rehearsal](https://mlsanity.com/paper/2403.01244/) (2024)
- [LFPT5: A Unified Framework for Lifelong Few-shot Language Learning Based on Prompt Tuning of T5](https://mlsanity.com/paper/2110.07298/) (2021)
- [Continual Learning for Text Classification with Information Disentanglement Based Regularization](https://mlsanity.com/paper/2104.05489/) (2021)
- [Towards Lifelong Learning of Large Language Models: A Survey](https://mlsanity.com/paper/2406.06391/) (2024)
- [InsCL: A Data-efficient Continual Learning Paradigm for Fine-tuning Large Language Models with Instructions](https://mlsanity.com/paper/2403.11435/) (2024)
- [From Language Modeling to Instruction Following: Understanding the Behavior Shift in LLMs after Instruction Tuning](https://mlsanity.com/paper/2310.00492/) (2023)
- [CITB: A Benchmark for Continual Instruction Tuning](https://mlsanity.com/paper/2310.14510/) (2023)
- [Investigating Catastrophic Forgetting During Continual Training for Neural Machine Translation](https://mlsanity.com/paper/2011.00678/) (2020)
- [Continual Named Entity Recognition without Catastrophic Forgetting](https://mlsanity.com/paper/2310.14541/) (2023)
- [MindLLM: Pre-training Lightweight Large Language Model from Scratch, Evaluations and Domain Applications](https://mlsanity.com/paper/2310.15777/) (2023)
- [Focus on Your Question! Interpreting and Mitigating Toxic CoT Problems in Commonsense Reasoning](https://mlsanity.com/paper/2402.18344/) (2024)
- [CorpusBrain++: A Continual Generative Pre-Training Framework for Knowledge-Intensive Language Tasks](https://mlsanity.com/paper/2402.16767/) (2024)
- [Word Matters: What Influences Domain Adaptation in Summarization?](https://mlsanity.com/paper/2406.14828/) (2024)

## Built on by (most cited first, 3 of 3)

- [Rethinking Transfer in Continual Learning: A Replay-Based Realisation](https://mlsanity.com/paper/2607.15587/) (2026)
- [KSOD: Knowledge Supplement for LLMs On Demand](https://mlsanity.com/paper/2503.07550/) (2025)
- [Can a Language Model Learn Facts Continually in Its Weights?](https://mlsanity.com/paper/2607.11020/) (2026)

## Abstract (arXiv)

Although substantial efforts have been made to mitigate catastrophic forgetting in continual learning, the intrinsic mechanisms are not well understood. In this work, we demonstrate the existence of "pseudo forgetting": the performance degradation on previous tasks is not attributed to a loss of capabilities, but rather to the failure of the instructions to activate the appropriate model abilities. We show that the model's performance on previous tasks can be restored through two simple interventions: (1) providing partial external correct rationale, and (2) appending semantically meaningless suffixes to the original instructions, to guide the generation of correct rationales. Through empirical analysis of the internal mechanisms governing rationale generation, we reveal that models exhibiting pseudo forgetting show reduced instruction dependence during rationale generation, leading to suboptimal activation of their inherent capabilities. Based on this insight, we propose Rationale-Guidance Difficulty based Replay (RGD-R) framework that dynamically allocates replay data based on the model's ability to correctly leverage the intrinsic capabilities. Experimental results demonstrate that RGD-R effectively mitigates pseudo forgetting while maintaining model plasticity.