Detecting and Pruning Prominent but Detrimental Neurons in Large Language Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909686552330240 |
|---|---|
| author | Ali, Ameen Katz, Shahar Wolf, Lior Titov, Ivan |
| author_facet | Ali, Ameen Katz, Shahar Wolf, Lior Titov, Ivan |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_09185 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Detecting and Pruning Prominent but Detrimental Neurons in Large Language Models Ali, Ameen Katz, Shahar Wolf, Lior Titov, Ivan Computation and Language Machine Learning 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. |
| title | Detecting and Pruning Prominent but Detrimental Neurons in Large Language Models |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2507.09185 |