PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference

• URL: http://arxiv.org/abs/2503.23274v1
• Date: Sun, 30 Mar 2025 01:47:23 GMT
• Title: PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference
• Authors: Weisheng Jin, Maojia Song, Tej Deep Pala, Yew Ken Chia, Amir Zadeh, Chuan Li, Soujanya Poria,
• Abstract summary: We propose PromptDistill, a training-free method that improves inference efficiency while preserving generation quality.<n>PromptDistill identifies and retains the most informative tokens by leveraging attention interactions in early layers, preserving their hidden states while reducing the computational burden in later layers.
• Score: 23.828853446344663
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: As large language models (LLMs) tackle increasingly complex tasks and longer documents, their computational and memory costs during inference become a major bottleneck. To address this, we propose PromptDistill, a novel, training-free method that improves inference efficiency while preserving generation quality. PromptDistill identifies and retains the most informative tokens by leveraging attention interactions in early layers, preserving their hidden states while reducing the computational burden in later layers. This allows the model to focus on essential contextual information without fully processing all tokens. Unlike previous methods such as H2O and SnapKV, which perform compression only after processing the entire input, or GemFilter, which selects a fixed portion of the initial prompt without considering contextual dependencies, PromptDistill dynamically allocates computational resources to the most relevant tokens while maintaining a global awareness of the input. Experiments using our method and baseline approaches with base models such as LLaMA 3.1 8B Instruct, Phi 3.5 Mini Instruct, and Qwen2 7B Instruct on benchmarks including LongBench, InfBench, and Needle in a Haystack demonstrate that PromptDistill significantly improves efficiency while having minimal impact on output quality compared to the original models. With a single-stage selection strategy, PromptDistill effectively balances performance and efficiency, outperforming prior methods like GemFilter, H2O, and SnapKV due to its superior ability to retain essential information. Specifically, compared to GemFilter, PromptDistill achieves an overall $1\%$ to $5\%$ performance improvement while also offering better time efficiency. Additionally, we explore multi-stage selection, which further improves efficiency while maintaining strong generation performance.

Related papers

• Efficient Token Compression for Vision Transformer with Spatial Information Preserved [59.79302182800274]
  Token compression is essential for reducing the computational and memory requirements of transformer models.<n>We propose an efficient and hardware-compatible token compression method called Prune and Merge.
  arXiv  Detail & Related papers  (2025-03-30T14:23:18Z)
• Achieving More with Less: Additive Prompt Tuning for Rehearsal-Free Class-Incremental Learning [76.32953653161417]
  Class-incremental learning enables models to learn new classes progressively while preserving knowledge of previously learned ones. Recent advances in this field have shifted towards parameter-efficient fine-tuning techniques. We present a novel prompt-based approach that addresses the limitation of current approaches.
  arXiv  Detail & Related papers  (2025-03-11T02:27:37Z)
• Efficient Prompt Compression with Evaluator Heads for Long-Context Transformer Inference [6.699062502635993]
  Long-context inputs are crucial for the effective utilization of large language models (LLMs) We propose an efficient, training-free prompt compression method that retains key information within compressed prompts. We develop Evaluator Head-based Prompt Compression, which enables LLMs to rapidly "skim through" input prompts.
  arXiv  Detail & Related papers  (2025-01-22T15:33:17Z)
• Compressing KV Cache for Long-Context LLM Inference with Inter-Layer Attention Similarity [24.118503938098307]
  Existing methods, including selective token retention and window-based attention, improve efficiency but risk discarding important tokens needed for future text generation.<n>We propose an approach that enhances LLM efficiency without token loss by reducing the memory and computational load of less important tokens, rather than discarding them.
  arXiv  Detail & Related papers  (2024-12-03T08:29:27Z)
• RefreshKV: Updating Small KV Cache During Long-form Generation [54.00118604124301]
  We propose a new inference method, RefreshKV, that flexibly alternates between full context attention and attention over a subset of input tokens during generation.<n>Applying our method to off-the-shelf LLMs achieves comparable speedup to eviction-based methods while improving performance for various long-form generation tasks.
  arXiv  Detail & Related papers  (2024-11-08T18:57:07Z)
• Training-Free Exponential Context Extension via Cascading KV Cache [49.608367376911694]
  We introduce a novel mechanism that leverages cascading sub-cache buffers to selectively retain the most relevant tokens. Our method reduces prefill stage latency by a factor of 6.8 when compared to flash attention on 1M tokens.
  arXiv  Detail & Related papers  (2024-06-24T03:59:17Z)
• CItruS: Chunked Instruction-aware State Eviction for Long Sequence Modeling [52.404072802235234]
  We introduce Chunked Instruction-aware State Eviction (CItruS), a novel modeling technique that integrates the attention preferences useful for a downstream task into the eviction process of hidden states. Our training-free method exhibits superior performance on long sequence comprehension and retrieval tasks over several strong baselines under the same memory budget.
  arXiv  Detail & Related papers  (2024-06-17T18:34:58Z)
• StreamFP: Learnable Fingerprint-guided Data Selection for Efficient Stream Learning [26.832497051355553]
  Stream Learning (SL) requires models that can quickly adapt to continuously evolving data.<n>Traditional rule-based data selection methods struggle to accommodate the dynamic nature of streaming data.<n>Recent approaches to handling changing data distributions face challenges that limit their effectiveness in fast-paced environments.
  arXiv  Detail & Related papers  (2024-06-11T10:46:41Z)
• REP: Resource-Efficient Prompting for Rehearsal-Free Continual Learning [23.92661395403251]
  Recent rehearsal-free methods, guided by prompts, excel in vision-related continual learning (CL) with drifting data but lack resource efficiency. We introduce Resource-Efficient Prompting (REP), which improves the computational and memory efficiency of prompt-based rehearsal-free methods. Our approach employs swift prompt selection to refine input data using a carefully provisioned model.
  arXiv  Detail & Related papers  (2024-06-07T09:17:33Z)
• Accelerating LLaMA Inference by Enabling Intermediate Layer Decoding via Instruction Tuning with LITE [62.13435256279566]
  Large Language Models (LLMs) have achieved remarkable performance across a wide variety of natural language tasks. However, their large size makes their inference slow and computationally expensive. We show that it enables these layers to acquire 'good' generation ability without affecting the generation ability of the final layer.
  arXiv  Detail & Related papers  (2023-10-28T04:07:58Z)
• Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers [29.319666323947708]
  We present a novel approach that dynamically prunes contextual information while preserving the model's expressiveness. Our method employs a learnable mechanism that determines which uninformative tokens can be dropped from the context. Our reference implementation achieves up to $2times$ increase in inference throughput and even greater memory savings.
  arXiv  Detail & Related papers  (2023-05-25T07:39:41Z)
• Efficient Few-Shot Object Detection via Knowledge Inheritance [62.36414544915032]
  Few-shot object detection (FSOD) aims at learning a generic detector that can adapt to unseen tasks with scarce training samples. We present an efficient pretrain-transfer framework (PTF) baseline with no computational increment. We also propose an adaptive length re-scaling (ALR) strategy to alleviate the vector length inconsistency between the predicted novel weights and the pretrained base weights.
  arXiv  Detail & Related papers  (2022-03-23T06:24:31Z)

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.