When Attention Collapses: How Degenerate Layers in LLMs Enable Smaller, Stronger Models

• URL: http://arxiv.org/abs/2404.08634v3
• Date: Sun, 08 Jun 2025 09:19:32 GMT
• Title: When Attention Collapses: How Degenerate Layers in LLMs Enable Smaller, Stronger Models
• Authors: Sunny Sanyal, Ravid Shwartz-Ziv, Alexandros G. Dimakis, Sujay Sanghavi,
• Abstract summary: Inheritune is a simple and effective training recipe for building smaller, more efficient language models.<n>We show that Inheritune trained models, despite having significantly fewer layers, can match or even outperform their larger counterparts.
• Score: 61.363259848264725
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) rely on the transformer architecture and its self-attention mechanism to deliver strong performance across tasks. However, we uncover a structural inefficiency in standard pre-trained decoder-style LLMs: in many of the deeper layers, attention matrices frequently collapse to near rank-one, single-column patterns. We refer to these underutilized components as lazy layers, which are redundant and computationally inefficient. To address this, we propose Inheritune, a simple and effective training recipe for building smaller, more efficient, and high performing language models. Inheritune initializes a compact model by inheriting the useful early layers from a larger pre-trained model, then progressively retrains and expands it. Our experiments across multiple models and datasets show that Inheritune trained models, despite having significantly fewer layers, can match or even outperform their larger counterparts. This approach yields compact, performant models and offers a practical path for efficient language model compression. Code is available at https://github.com/sanyalsunny111/LLM-Inheritune

Related papers

• LESA: Learnable LLM Layer Scaling-Up [57.0510934286449]
  Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive.<n>Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones.<n>We propose textbfLESA, a novel learnable method for depth scaling-up.
  arXiv  Detail & Related papers  (2025-02-19T14:58:48Z)
• Single Parent Family: A Spectrum of Family Members from a Single Pre-Trained Foundation Model [20.054342930450055]
  This paper introduces a novel method of Progressive Low Rank Decomposition (PLRD) tailored for the compression of large language models. PLRD allows for significant reductions in computational overhead and energy consumption. Our findings suggest that PLRD could set a new standard for the efficient scaling of LLMs.
  arXiv  Detail & Related papers  (2024-06-28T15:27:57Z)
• Evolving Subnetwork Training for Large Language Models [19.54861230097017]
  We propose a novel training paradigm: Evolving Subnetwork Training (EST) EST samplesworks from the layers of the large language model and from commonly used modules within each layer. We apply EST to train GPT2 model and TinyLlama model, resulting in 26.7% FLOPs saving for GPT2 and 25.0% for TinyLlama without an increase in loss on the pre-training dataset.
  arXiv  Detail & Related papers  (2024-06-11T05:44:56Z)
• ShortGPT: Layers in Large Language Models are More Redundant Than You Expect [38.148626520751385]
  We show that many layers of Large Language Models (LLMs) exhibit high similarity, and some layers play a negligible role in network functionality. We propose a straightforward pruning approach: layer removal, in which we directly delete the redundant layers. Experiments demonstrate that our method, which we call ShortGPT, significantly outperforms previous state-of-the-art (SOTA) methods in model pruning.
  arXiv  Detail & Related papers  (2024-03-06T17:04:18Z)
• Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning [52.29522018586365]
  We study structured pruning as an effective means to develop smaller LLMs from pre-trained, larger models. Our approach employs two key techniques: (1) targeted structured pruning, which prunes a larger model to a specified target shape by removing layers, heads, and intermediate and hidden dimensions in an end-to-end manner, and (2) dynamic batch loading, which dynamically updates the composition of sampled data in each training batch based on varying losses across different domains.
  arXiv  Detail & Related papers  (2023-10-10T15:13:30Z)
• Enhancing Cross-Category Learning in Recommendation Systems with Multi-Layer Embedding Training [2.4862527485819186]
  Multi-layer embeddings training (MLET) trains embeddings using factorization of the embedding layer, with an inner dimension higher than the target embedding dimension. MLET consistently produces better models, especially for rare items. At constant model quality, MLET allows embedding dimension, and model size, reduction by up to 16x, and 5.8x on average.
  arXiv  Detail & Related papers  (2023-09-27T09:32:10Z)
• Masked Image Modeling with Local Multi-Scale Reconstruction [54.91442074100597]
  Masked Image Modeling (MIM) achieves outstanding success in self-supervised representation learning. Existing MIM models conduct reconstruction task only at the top layer of encoder. We design local multi-scale reconstruction, where the lower and upper layers reconstruct fine-scale and coarse-scale supervision signals respectively.
  arXiv  Detail & Related papers  (2023-03-09T13:42:04Z)
• Training Trajectories of Language Models Across Scales [99.38721327771208]
  Scaling up language models has led to unprecedented performance gains. How do language models of different sizes learn during pre-training? Why do larger language models demonstrate more desirable behaviors?
  arXiv  Detail & Related papers  (2022-12-19T19:16:29Z)
• LV-BERT: Exploiting Layer Variety for BERT [85.27287501885807]
  We introduce convolution into the layer type set, which is experimentally found beneficial to pre-trained models. We then adopt an evolutionary algorithm guided by pre-training accuracy to find the optimal architecture. LV-BERT model obtained by our method outperforms BERT and its variants on various downstream tasks.
  arXiv  Detail & Related papers  (2021-06-22T13:20:14Z)
• Top-KAST: Top-K Always Sparse Training [50.05611544535801]
  We propose Top-KAST, a method that preserves constant sparsity throughout training. We show that it performs comparably to or better than previous works when training models on the established ImageNet benchmark. In addition to our ImageNet results, we also demonstrate our approach in the domain of language modeling.
  arXiv  Detail & Related papers  (2021-06-07T11:13:05Z)
• Training with Multi-Layer Embeddings for Model Reduction [0.9046327456472286]
  We introduce a multi-layer embedding training architecture that trains embeddings via a sequence of linear layers. We show that it allows reducing d by 4-8X, with a corresponding improvement in memory footprint, at given model accuracy.
  arXiv  Detail & Related papers  (2020-06-10T02:47:40Z)
• Train Large, Then Compress: Rethinking Model Size for Efficient Training and Inference of Transformers [94.43313684188819]
  We study the impact of model size in this setting, focusing on Transformer models for NLP tasks that are limited by compute. We first show that even though smaller Transformer models execute faster per iteration, wider and deeper models converge in significantly fewer steps. This leads to an apparent trade-off between the training efficiency of large Transformer models and the inference efficiency of small Transformer models.
  arXiv  Detail & Related papers  (2020-02-26T21:17:13Z)

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.