Unveiling and Addressing Pseudo Forgetting in Large Language Models

• URL: http://arxiv.org/abs/2411.11932v2
• Date: Mon, 09 Jun 2025 02:09:36 GMT
• Title: Unveiling and Addressing Pseudo Forgetting in Large Language Models
• Authors: Huashan Sun, Yizhe Yang, Yinghao Li, Jiawei Li, Yang Gao,
• Abstract summary: We show that 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.<n>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.
• Score: 17.888328120571245
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization [48.91511514636768]
  We present Length-Adaptive Policy Optimization (LAPO), a framework that transforms reasoning length control from an external constraint into an intrinsic model capability.<n>LAPO enables models to internalize an understanding of appropriate reasoning depth through a two-stage reinforcement learning process.<n> Experiments on mathematical reasoning benchmarks demonstrate that LAPO reduces token usage by up to 40.9% while improving accuracy by 2.3%.
  arXiv  Detail & Related papers  (2025-07-21T16:14:41Z)
• Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancement [101.77467538102924]
  Large reasoning models (LRMs) exhibit overthinking, which hinders efficiency and inflates inference cost.<n>We propose two lightweight methods to enhance LRM efficiency.<n>First, we introduce Efficiency Steering, a training-free activation steering technique that modulates reasoning behavior via a single direction.<n>Second, we develop Self-Rewarded Efficiency RL, a reinforcement learning framework that dynamically balances task accuracy and brevity.
  arXiv  Detail & Related papers  (2025-06-18T17:18:12Z)
• Rationales Are Not Silver Bullets: Measuring the Impact of Rationales on Model Performance and Reliability [70.4107059502882]
  Training language models with rationales augmentation has been shown to be beneficial in many existing works.<n>We conduct comprehensive investigations to thoroughly inspect the impact of rationales on model performance.
  arXiv  Detail & Related papers  (2025-05-30T02:39:37Z)
• ASIDE: Architectural Separation of Instructions and Data in Language Models [87.16417239344285]
  We propose a method, ASIDE, that allows the model to clearly separate between instructions and data on the level of embeddings. ASIDE applies a fixed rotation to the embeddings of data tokens, thus creating distinct representations of instructions and data tokens without introducing any additional parameters. We demonstrate the effectiveness of our method by instruct-tuning LLMs with ASIDE and showing (1) highly increased instruction-data separation scores without a loss in model capabilities and (2) competitive results on prompt injection benchmarks, even without dedicated safety training.
  arXiv  Detail & Related papers  (2025-03-13T17:17:17Z)
• Self-rewarding correction for mathematical reasoning [19.480508580498103]
  We study self-rewarding reasoning large language models (LLMs)<n>LLMs can simultaneously generate step-by-step reasoning and evaluate the correctness of their outputs during the inference time-without external feedback.<n>We propose a two-staged algorithmic framework for constructing self-rewarding reasoning models using only self-generated data.
  arXiv  Detail & Related papers  (2025-02-26T23:01:16Z)
• Bridging Internal Probability and Self-Consistency for Effective and Efficient LLM Reasoning [53.25336975467293]
  We present the first theoretical error decomposition analysis of methods such as perplexity and self-consistency.<n>Our analysis reveals a fundamental trade-off: perplexity methods suffer from substantial model error due to the absence of a proper consistency function.<n>We propose Reasoning-Pruning Perplexity Consistency (RPC), which integrates perplexity with self-consistency, and Reasoning Pruning, which eliminates low-probability reasoning paths.
  arXiv  Detail & Related papers  (2025-02-01T18:09:49Z)
• B-STaR: Monitoring and Balancing Exploration and Exploitation in Self-Taught Reasoners [18.960920426485163]
  Self-improvement has emerged as a primary method for enhancing performance.<n>We identify and propose methods to monitor two pivotal factors in this iterative process.<n>We introduce B-STaR, a Self-Taught Reasoning framework that adjusts configurations across iterations to balance exploration and exploitation.
  arXiv  Detail & Related papers  (2024-12-23T03:58:34Z)
• RESTOR: Knowledge Recovery through Machine Unlearning [71.75834077528305]
  Large language models trained on web-scale corpora can memorize undesirable datapoints. Many machine unlearning algorithms have been proposed that aim to erase' these datapoints. We propose the RESTOR framework for machine unlearning, which evaluates the ability of unlearning algorithms to perform targeted data erasure.
  arXiv  Detail & Related papers  (2024-10-31T20:54:35Z)
• Idempotent Unsupervised Representation Learning for Skeleton-Based Action Recognition [13.593511876719367]
  We propose a novel skeleton-based idempotent generative model (IGM) for unsupervised representation learning. Our experiments on benchmark datasets, NTU RGB+D and PKUMMD, demonstrate the effectiveness of our proposed method.
  arXiv  Detail & Related papers  (2024-10-27T06:29:04Z)
• Unlearning with Control: Assessing Real-world Utility for Large Language Model Unlearning [97.2995389188179]
  Recent research has begun to approach large language models (LLMs) unlearning via gradient ascent (GA) Despite their simplicity and efficiency, we suggest that GA-based methods face the propensity towards excessive unlearning. We propose several controlling methods that can regulate the extent of excessive unlearning.
  arXiv  Detail & Related papers  (2024-06-13T14:41:00Z)
• Improving Language Model Reasoning with Self-motivated Learning [60.779625789039486]
  textitSelf-motivated Learning framework motivates the model itself to automatically generate rationales on existing datasets. We train a reward model with the rank to evaluate the quality of rationales, and improve the performance of reasoning through reinforcement learning.
  arXiv  Detail & Related papers  (2024-04-10T14:05:44Z)
• Goldfish: An Efficient Federated Unlearning Framework [3.956103498302838]
  Goldfish is a new framework for machine unlearning algorithms. It comprises four modules: basic model, loss function, optimization, and extension. To address the challenge of low validity in existing machine unlearning algorithms, we propose a novel loss function.
  arXiv  Detail & Related papers  (2024-04-04T03:29:41Z)
• STAR: Constraint LoRA with Dynamic Active Learning for Data-Efficient Fine-Tuning of Large Language Models [21.929902181609936]
  We propose a novel approach to integrate uncertainty-based active learning and LoRA. For the uncertainty gap, we introduce a dynamic uncertainty measurement that combines the uncertainty of the base model and the uncertainty of the full model. For poor model calibration, we incorporate the regularization method during LoRA training to keep the model from being over-confident.
  arXiv  Detail & Related papers  (2024-03-02T10:38:10Z)
• An Information Theoretic Approach to Machine Unlearning [45.600917449314444]
  Key challenge in unlearning is forgetting the necessary data in a timely manner, while preserving model performance. In this work, we address the zero-shot unlearning scenario, whereby an unlearning algorithm must be able to remove data given only a trained model and the data to be forgotten. We derive a simple but principled zero-shot unlearning method based on the geometry of the model.
  arXiv  Detail & Related papers  (2024-02-02T13:33:30Z)
• EsaCL: Efficient Continual Learning of Sparse Models [10.227171407348326]
  Key challenge in the continual learning setting is to efficiently learn a sequence of tasks without forgetting how to perform previously learned tasks. We propose a new method for efficient continual learning of sparse models (EsaCL) that can automatically prune redundant parameters without adversely impacting the model's predictive power.
  arXiv  Detail & Related papers  (2024-01-11T04:59:44Z)
• Understanding, Predicting and Better Resolving Q-Value Divergence in Offline-RL [86.0987896274354]
  We first identify a fundamental pattern, self-excitation, as the primary cause of Q-value estimation divergence in offline RL. We then propose a novel Self-Excite Eigenvalue Measure (SEEM) metric to measure the evolving property of Q-network at training. For the first time, our theory can reliably decide whether the training will diverge at an early stage.
  arXiv  Detail & Related papers  (2023-10-06T17:57:44Z)
• Towards Robust Continual Learning with Bayesian Adaptive Moment Regularization [51.34904967046097]
  Continual learning seeks to overcome the challenge of catastrophic forgetting, where a model forgets previously learnt information. We introduce a novel prior-based method that better constrains parameter growth, reducing catastrophic forgetting. Results show that BAdam achieves state-of-the-art performance for prior-based methods on challenging single-headed class-incremental experiments.
  arXiv  Detail & Related papers  (2023-09-15T17:10:51Z)
• Machine Unlearning Methodology base on Stochastic Teacher Network [33.763901254862766]
  "Right to be forgotten" grants data owners the right to actively withdraw data that has been used for model training. Existing machine unlearning methods have been found to be ineffective in quickly removing knowledge from deep learning models. This paper proposes using a network as a teacher to expedite the mitigation of the influence caused by forgotten data on the model.
  arXiv  Detail & Related papers  (2023-08-28T06:05:23Z)
• Machine Unlearning of Features and Labels [72.81914952849334]
  We propose first scenarios for unlearning and labels in machine learning models. Our approach builds on the concept of influence functions and realizes unlearning through closed-form updates of model parameters.
  arXiv  Detail & Related papers  (2021-08-26T04:42:24Z)
• Distilling Interpretable Models into Human-Readable Code [71.11328360614479]
  Human-readability is an important and desirable standard for machine-learned model interpretability. We propose to train interpretable models using conventional methods, and then distill them into concise, human-readable code. We describe a piecewise-linear curve-fitting algorithm that produces high-quality results efficiently and reliably across a broad range of use cases.
  arXiv  Detail & Related papers  (2021-01-21T01:46:36Z)
• Progressive Self-Guided Loss for Salient Object Detection [102.35488902433896]
  We present a progressive self-guided loss function to facilitate deep learning-based salient object detection in images. Our framework takes advantage of adaptively aggregated multi-scale features to locate and detect salient objects effectively.
  arXiv  Detail & Related papers  (2021-01-07T07:33:38Z)
• Automatic Recall Machines: Internal Replay, Continual Learning and the Brain [104.38824285741248]
  Replay in neural networks involves training on sequential data with memorized samples, which counteracts forgetting of previous behavior caused by non-stationarity. We present a method where these auxiliary samples are generated on the fly, given only the model that is being trained for the assessed objective. Instead the implicit memory of learned samples within the assessed model itself is exploited.
  arXiv  Detail & Related papers  (2020-06-22T15:07:06Z)

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.