ALPS: Attention Localization and Pruning Strategy for Efficient Alignment of Large Language Models

• URL: http://arxiv.org/abs/2505.18799v4
• Date: Wed, 18 Jun 2025 05:56:01 GMT
• Title: ALPS: Attention Localization and Pruning Strategy for Efficient Alignment of Large Language Models
• Authors: Hao Chen, Haoze Li, Zhiqing Xiao, Lirong Gao, Qi Zhang, Xiaomeng Hu, Ningtao Wang, Xing Fu, Junbo Zhao,
• Abstract summary: We propose an efficient algorithm that localizes the most task-sensitive attention heads and prunes them by restricting attention training updates to these heads.<n> Experimental results demonstrate that our method activates only 10% of attention parameters during fine-tuning while achieving a 2% performance improvement over baselines on three tasks.
• Score: 14.657194214702473
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Aligning general-purpose large language models (LLMs) to downstream tasks often incurs significant training adjustment costs. Prior research has explored various avenues to enhance alignment efficiency, primarily through minimal-data training or data-driven activations to identify key attention heads. However, these approaches inherently introduce data dependency, which hinders generalization and reusability. To address this issue and enhance model alignment efficiency, we propose the Attention Localization and Pruning Strategy (ALPS), an efficient algorithm that localizes the most task-sensitive attention heads and prunes by restricting attention training updates to these heads, thereby reducing alignment costs. Experimental results demonstrate that our method activates only 10% of attention parameters during fine-tuning while achieving a 2% performance improvement over baselines on three tasks. Moreover, the identified task-specific heads are transferable across datasets and mitigate knowledge forgetting. Our work and findings provide a novel perspective on efficient LLM alignment. The code is available at https://github.com/VoiceBeer/ALPS.

Related papers

• Optimising Language Models for Downstream Tasks: A Post-Training Perspective [0.0]
  Language models (LMs) have demonstrated remarkable capabilities in NLP.<n>But adapting them efficiently and robustly to specific tasks remains challenging.<n>This thesis proposes a series of methods to better adapt LMs to downstream applications.
  arXiv  Detail & Related papers  (2025-06-26T00:49:35Z)
• Improving Data Efficiency for LLM Reinforcement Fine-tuning Through Difficulty-targeted Online Data Selection and Rollout Replay [61.823835392216544]
  Reinforcement learning (RL) has become an effective approach for fine-tuning large language models (LLMs)<n>We propose two techniques to improve data efficiency in LLM RL fine-tuning: difficulty-targeted online data selection and rollout replay.<n>Our method reduces RL fine-tuning time by 25% to 65% to reach the same level of performance as the original GRPO algorithm.
  arXiv  Detail & Related papers  (2025-06-05T17:55:43Z)
• KARE-RAG: Knowledge-Aware Refinement and Enhancement for RAG [63.82127103851471]
  Retrieval-Augmented Generation (RAG) enables large language models to access broader knowledge sources.<n>We demonstrate that enhancing generative models' capacity to process noisy content is equally critical for robust performance.<n>We present KARE-RAG, which improves knowledge utilization through three key innovations.
  arXiv  Detail & Related papers  (2025-06-03T06:31:17Z)
• Enhancing Training Data Attribution with Representational Optimization [57.61977909113113]
  Training data attribution methods aim to measure how training data impacts a model's predictions.<n>We propose AirRep, a representation-based approach that closes this gap by learning task-specific and model-aligned representations explicitly for TDA.<n>AirRep introduces two key innovations: a trainable encoder tuned for attribution quality, and an attention-based pooling mechanism that enables accurate estimation of group-wise influence.
  arXiv  Detail & Related papers  (2025-05-24T05:17:53Z)
• R-Sparse: Rank-Aware Activation Sparsity for Efficient LLM Inference [77.47238561728459]
  R-Sparse is a training-free activation sparsity approach capable of achieving high sparsity levels in advanced LLMs.<n> Experiments on Llama-2/3 and Mistral models across ten diverse tasks demonstrate that R-Sparse achieves comparable performance at 50% model-level sparsity.
  arXiv  Detail & Related papers  (2025-04-28T03:30:32Z)
• Question Tokens Deserve More Attention: Enhancing Large Language Models without Training through Step-by-Step Reading and Question Attention Recalibration [0.36561146074362716]
  Large Language Models (LLMs) often struggle with tasks that require a deep understanding of complex questions.<n>This work investigates the limitations of current LLMs in question comprehension.<n>We propose a family of prompt-based strategies that guide LLMs to incrementally process question tokens and align their reasoning with the input structure.
  arXiv  Detail & Related papers  (2025-04-13T02:10:18Z)
• Learning Task Representations from In-Context Learning [73.72066284711462]
  Large language models (LLMs) have demonstrated remarkable proficiency in in-context learning.<n>We introduce an automated formulation for encoding task information in ICL prompts as a function of attention heads.<n>We show that our method's effectiveness stems from aligning the distribution of the last hidden state with that of an optimally performing in-context-learned model.
  arXiv  Detail & Related papers  (2025-02-08T00:16:44Z)
• Aligning Instruction Tuning with Pre-training [81.4748965653345]
  We propose Aligning Instruction Tuning with Pre-training (AITP) to align instruction tuning with pre-training distributions.<n>We show consistent performance improvements with AITP on three fully open large language models (LLMs) across eight benchmarks.
  arXiv  Detail & Related papers  (2025-01-16T08:27:40Z)
• Beyond Task Vectors: Selective Task Arithmetic Based on Importance Metrics [0.0]
  This paper introduces textbfunderlineSelective textbfunderlineTask textbfunderlineArithmetic underlinetextbf(STA), a training-free framework designed to enhance multi-task performance through task-specific parameter fusion. Experimental results demonstrate that STA achieves superior multi-task performance across benchmarks and excellent performance in task forgetting.
  arXiv  Detail & Related papers  (2024-11-25T06:59:16Z)
• SEEKR: Selective Attention-Guided Knowledge Retention for Continual Learning of Large Language Models [27.522743690956315]
  We propose a SElective attEntion-guided Knowledge Retention method (SEEKR) for data-efficient replay-based continual learning of large language models (LLMs) SEEKR performs attention distillation on the selected attention heads for finer-grained knowledge retention. Experimental results on two continual learning benchmarks for LLMs demonstrate the superiority of SEEKR over the existing methods on both performance and efficiency.
  arXiv  Detail & Related papers  (2024-11-09T13:02:36Z)
• Enhancing Robustness of Vision-Language Models through Orthogonality Learning and Self-Regularization [77.62516752323207]
  We introduce an orthogonal fine-tuning method for efficiently fine-tuning pretrained weights and enabling enhanced robustness and generalization. A self-regularization strategy is further exploited to maintain the stability in terms of zero-shot generalization of VLMs, dubbed OrthSR. For the first time, we revisit the CLIP and CoOp with our method to effectively improve the model on few-shot image classficiation scenario.
  arXiv  Detail & Related papers  (2024-07-11T10:35:53Z)
• Fully Fine-tuned CLIP Models are Efficient Few-Shot Learners [8.707819647492467]
  We explore capturing the task-specific information via meticulous refinement of entire Vision-Language Models (VLMs) To mitigate these issues, we propose a framework named CLIP-CITE via designing a discriminative visual-text task.
  arXiv  Detail & Related papers  (2024-07-04T15:22:54Z)
• Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration [15.36841874118801]
  We aim to provide a more profound understanding of the existence of attention sinks within large language models (LLMs) We propose a training-free Attention Technique (ACT) that automatically optimize the attention distributions on the fly during inference in an input-adaptive manner. ACT achieves an average improvement of up to 7.30% in accuracy across different datasets when applied to Llama-30B.
  arXiv  Detail & Related papers  (2024-06-22T07:00:43Z)
• HiFi: High-Information Attention Heads Hold for Parameter-Efficient Model Adaptation [0.8409934249521909]
  We propose a parameter-efficient fine-tuning method HiFi, that is, only the highly informative and strongly correlated attention heads for the specific task are fine-tuned. We first model the relationship between heads into a graph from two perspectives of information richness and correlation, and then apply PageRank algorithm to determine the relative importance of each head. Experiments on the GLUE benchmark demonstrate the effectiveness of our method, and show that HiFi obtains state-of-the-art performance over the prior baselines.
  arXiv  Detail & Related papers  (2023-05-08T09:31:13Z)
• Exploring and Predicting Transferability across NLP Tasks [115.6278033699853]
  We study the transferability between 33 NLP tasks across three broad classes of problems. Our results show that transfer learning is more beneficial than previously thought. We also develop task embeddings that can be used to predict the most transferable source tasks for a given target task.
  arXiv  Detail & Related papers  (2020-05-02T09:39:36Z)

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.