Detecting and Pruning Prominent but Detrimental Neurons in Large Language Models

• URL: http://arxiv.org/abs/2507.09185v1
• Date: Sat, 12 Jul 2025 08:10:10 GMT
• Title: Detecting and Pruning Prominent but Detrimental Neurons in Large Language Models
• Authors: Ameen Ali, Shahar Katz, Lior Wolf, Ivan Titov,
• Abstract summary: Large language models (LLMs) often develop learned mechanisms specialized to specific datasets.<n>We introduce a fine-tuning approach designed to enhance generalization by identifying and pruning neurons associated with dataset-specific mechanisms.<n>Our method employs Integrated Gradients to quantify each neuron's influence on high-confidence predictions, pinpointing those that disproportionately contribute to dataset-specific performance.
• Score: 68.57424628540907
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) often develop learned mechanisms specialized to specific datasets, such as reliance on domain-specific correlations, which yield high-confidence predictions without generalizable reasoning. While beneficial in one setting, these dataset-specific mechanisms typically degrade performance when models encounter novel tasks or distributions. In this work, we introduce a fine-tuning approach designed to enhance generalization by identifying and pruning neurons associated with dataset-specific mechanisms in transformer-based LLMs. Our method employs Integrated Gradients to quantify each neuron's influence on high-confidence predictions, pinpointing those that disproportionately contribute to dataset-specific performance without supporting robust, transferable reasoning. Selectively pruning these neurons compels the model to depend on generalizable representations. Evaluated across multiple-choice benchmarks, our pruning-based fine-tuning significantly enhances performance, surpassing prior (non-pruning) adaptation methods.

Related papers

• Why Generate When You Can Transform? Unleashing Generative Attention for Dynamic Recommendation [9.365893765448366]
  Sequential Recommendation (SR) focuses on personalizing user experiences by predicting future preferences based on historical interactions.<n> Transformer models, with their attention mechanisms, have become the dominant architecture in SR tasks.<n>We introduce two generative attention models for SR, each grounded in the principles of Variational Autoencoders (VAE) and Diffusion Models (DMs)
  arXiv  Detail & Related papers  (2025-08-04T04:33:26Z)
• Information-theoretic Quantification of High-order Feature Effects in Classification Problems [0.19791587637442676]
  We present an information-theoretic extension of the High-order interactions for Feature importance (Hi-Fi) method.<n>Our framework decomposes feature contributions into unique, synergistic, and redundant components.<n>Results indicate that the proposed estimator accurately recovers theoretical and expected findings.
  arXiv  Detail & Related papers  (2025-07-06T11:50:30Z)
• Single Domain Generalization with Model-aware Parametric Batch-wise Mixup [22.709796153794507]
  Single Domain Generalization remains a formidable challenge in the field of machine learning.<n>We propose a novel data augmentation approach, named as Model-aware Parametric Batch-wise Mixup.<n>By exploiting inter-feature correlations, the parameterized mixup generator introduces additional versatility in combining features across a batch of instances.
  arXiv  Detail & Related papers  (2025-02-22T03:45:18Z)
• Influence Functions for Scalable Data Attribution in Diffusion Models [52.92223039302037]
  Diffusion models have led to significant advancements in generative modelling.<n>Yet their widespread adoption poses challenges regarding data attribution and interpretability.<n>We develop an influence functions framework to address these challenges.
  arXiv  Detail & Related papers  (2024-10-17T17:59:02Z)
• Regularized Neural Ensemblers [55.15643209328513]
  In this study, we explore employing regularized neural networks as ensemble methods.<n>Motivated by the risk of learning low-diversity ensembles, we propose regularizing the ensembling model by randomly dropping base model predictions.<n>We demonstrate this approach provides lower bounds for the diversity within the ensemble, reducing overfitting and improving generalization capabilities.
  arXiv  Detail & Related papers  (2024-10-06T15:25:39Z)
• GenBench: A Benchmarking Suite for Systematic Evaluation of Genomic Foundation Models [56.63218531256961]
  We introduce GenBench, a benchmarking suite specifically tailored for evaluating the efficacy of Genomic Foundation Models. GenBench offers a modular and expandable framework that encapsulates a variety of state-of-the-art methodologies. We provide a nuanced analysis of the interplay between model architecture and dataset characteristics on task-specific performance.
  arXiv  Detail & Related papers  (2024-06-01T08:01:05Z)
• Subject-specific Deep Neural Networks for Count Data with High-cardinality Categorical Features [1.2289361708127877]
  We propose a novel hierarchical likelihood learning framework for introducing gamma random effects into a Poisson deep neural network. The proposed method simultaneously yields maximum likelihood estimators for fixed parameters and best unbiased predictors for random effects. State-of-the-art network architectures can be easily implemented into the proposed h-likelihood framework.
  arXiv  Detail & Related papers  (2023-10-18T01:54:48Z)
• HyperImpute: Generalized Iterative Imputation with Automatic Model Selection [77.86861638371926]
  We propose a generalized iterative imputation framework for adaptively and automatically configuring column-wise models. We provide a concrete implementation with out-of-the-box learners, simulators, and interfaces.
  arXiv  Detail & Related papers  (2022-06-15T19:10:35Z)
• On the Generalization and Adaption Performance of Causal Models [99.64022680811281]
  Differentiable causal discovery has proposed to factorize the data generating process into a set of modules. We study the generalization and adaption performance of such modular neural causal models. Our analysis shows that the modular neural causal models outperform other models on both zero and few-shot adaptation in low data regimes.
  arXiv  Detail & Related papers  (2022-06-09T17:12:32Z)
• Supervised Autoencoders Learn Robust Joint Factor Models of Neural Activity [2.8402080392117752]
  neuroscience applications collect high-dimensional predictors' corresponding to brain activity in different regions along with behavioral outcomes. Joint factor models for the predictors and outcomes are natural, but maximum likelihood estimates of these models can struggle in practice when there is model misspecification. We propose an alternative inference strategy based on supervised autoencoders; rather than placing a probability distribution on the latent factors, we define them as an unknown function of the high-dimensional predictors.
  arXiv  Detail & Related papers  (2020-04-10T19:31:57Z)

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.