SHRP: Specialized Head Routing and Pruning for Efficient Encoder Compression

• URL: http://arxiv.org/abs/2512.20635v1
• Date: Wed, 03 Dec 2025 05:03:55 GMT
• Title: SHRP: Specialized Head Routing and Pruning for Efficient Encoder Compression
• Abstract summary: Transformer encoders are widely deployed in large-scale web services for natural language understanding tasks.<n>These limitations stem largely from architectural redundancy, particularly in the attention module.<n>We propose SHRP, a novel structured pruning framework that automatically identifies and removes redundant attention heads.
• Score: 11.234631528609603
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Transformer encoders are widely deployed in large-scale web services for natural language understanding tasks such as text classification, semantic retrieval, and content ranking. However, their high inference latency and memory consumption pose significant challenges for real-time serving and scalability. These limitations stem largely from architectural redundancy, particularly in the attention module. The inherent parameter redundancy of the attention mechanism, coupled with the fact that its attention heads operate with a degree of independence, makes it particularly amenable to structured model compression. In this paper, we propose SHRP (Specialized Head Routing and Pruning), a novel structured pruning framework that automatically identifies and removes redundant attention heads while preserving most of the model's accuracy and compatibility. SHRP introduces Expert Attention, a modular design that treats each attention head as an independent expert, followed by a lightweight shared expander feed-forward network that refines their outputs. The framework employs a unified Top-1 usage-driven mechanism to jointly perform dynamic routing during training and deterministic pruning at deployment. Experimental results on the GLUE benchmark using a BERT-base encoder show that SHRP achieves 93% of the original model accuracy while reducing parameters by 48 percent. Under an extreme compression scenario where 11/12 of the layers are pruned, the model still maintains 84% accuracy and delivers a 4.2x throughput gain while reducing computation to as low as 11.5 percent of the original FLOPs, demonstrating its practical utility for large-scale and latency-sensitive web deployments.

Related papers

• QuantSR+: Pushing the Limit of Quantized Image Super-Resolution Networks [63.04549067734431]
  Low-bit quantization is widely used to compress super-resolution (SR) models.<n>However, when SR models are pushed to ultra-low precision (2-4 bits), performance can drop sharply.<n>We propose QuantSR+, a unified framework that improves quantization operators, network design, and training optimization.
  arXiv  Detail & Related papers  (2026-05-21T11:38:52Z)
• Elastic Spiking Transformers for Efficient Gesture Understanding [6.52225846892392]
  Spiking Neural Networks (SNNs) offer energy-efficient processing of event-based sensor data for healthcare applications.<n>We introduce the Elastic Spiking Transformer, a runtime-adaptive architecture that brings elasticity into the spiking paradigm.<n>We evaluate the approach on CIFAR10/100, CIFAR10-DVS, and the EHWGesture clinical gesture understanding dataset.
  arXiv  Detail & Related papers  (2026-05-04T12:35:52Z)
• SMR-Net:Robot Snap Detection Based on Multi-Scale Features and Self-Attention Network [0.0]
  Traditional visual methods suffer from poor robustness and large localization errors when handling complex scenarios.<n>This paper proposes SMR-Net, a self-attention-based multi-scale object detection algorithm.<n> Experimental results on Type A and Type B snap datasets show SMR-Net outperforms traditional Faster R-CNN significantly.
  arXiv  Detail & Related papers  (2026-03-01T10:28:01Z)
• CompilerKV: Risk-Adaptive KV Compression via Offline Experience Compilation [16.540537092341737]
  We propose CompilerKV, a risk-adaptive and head-aware compression framework for large language models.<n>CoherentKV compiles offline experience into reusable decision tables for prefill-only deployment.<n>Experiments on LongBench show CompilerKV dominates SOTA methods under a 512-token budget, recovering 97.7% of FullKV performance while achieving up to +5.2 points gain over the strongest competitor.
  arXiv  Detail & Related papers  (2026-02-09T14:07:55Z)
• Towards Efficient Agents: A Co-Design of Inference Architecture and System [66.59916327634639]
  This paper presents AgentInfer, a unified framework for end-to-end agent acceleration.<n>We decompose the problem into four synergistic components: AgentCollab, AgentSched, AgentSAM, and AgentCompress.<n>Experiments on the BrowseComp-zh and DeepDiver benchmarks demonstrate that through the synergistic collaboration of these methods, AgentInfer reduces ineffective token consumption by over 50%.
  arXiv  Detail & Related papers  (2025-12-20T12:06:13Z)
• Compression is Routing: Reconstruction Error as an Intrinsic Signal for Modular Language Models [0.0]
  This paper builds upon the premise that Compression is Intelligence''<n>It proposes a novel architectural philosophy: Compression is Routing.<n>It offers a new perspective on VRAM compression'' for handling ultra-long contexts.
  arXiv  Detail & Related papers  (2025-12-18T09:02:03Z)
• Knowledge-Informed Neural Network for Complex-Valued SAR Image Recognition [51.03674130115878]
  We introduce the Knowledge-Informed Neural Network (KINN), a lightweight framework built upon a novel "compression-aggregation-compression" architecture.<n>KINN establishes a state-of-the-art in parameter-efficient recognition, offering exceptional generalization in data-scarce and out-of-distribution scenarios.
  arXiv  Detail & Related papers  (2025-10-23T07:12:26Z)
• From Local to Global: Revisiting Structured Pruning Paradigms for Large Language Models [27.774067682004745]
  GISP-Global Iterative Structured Pruning removes attention heads and channels using first-order, loss-based important weights aggregated at the structure level with block-wise normalization.<n>An iterative schedule, rather than one-shot pruning, stabilizes accuracy at higher sparsity and mitigates perplexity collapse without requiring intermediate fine-tuning.<n>Because importance is defined by a model-level loss, GISP naturally supports task-specific objectives; we instantiate perplexity for language modeling and a margin-based objective for decision-style tasks.
  arXiv  Detail & Related papers  (2025-10-20T19:04:09Z)
• MC#: Mixture Compressor for Mixture-of-Experts Large Models [86.64315380917827]
  Mixture-of-Experts (MoE) effectively scales large language models (LLMs) and vision-language models (VLMs) by increasing capacity through sparse activation.<n>We propose MC# (Mixture-Compressor-sharp), a framework that combines static quantization and dynamic expert pruning.
  arXiv  Detail & Related papers  (2025-10-13T03:12:46Z)
• Flexible Automatic Identification and Removal (FAIR)-Pruner: An Efficient Neural Network Pruning Method [11.575879702610914]
  This paper proposes the Flexible Automatic Identification and Removal (FAIR)-Pruner, a novel method for neural network structured pruning.<n>A major advantage of FAIR-Pruner lies in its capacity to automatically determine the layer-wise pruning rates, which yields a more efficient subnetwork structure.<n>With utilization scores and reconstruction errors, users can flexibly obtain pruned models under different pruning ratios.
  arXiv  Detail & Related papers  (2025-08-04T10:59:07Z)
• Application-Specific Component-Aware Structured Pruning of Deep Neural Networks via Soft Coefficient Optimization [1.6874375111244326]
  It remains critical to ensure that application-specific performance characteristics are preserved during compression.<n>In structured pruning, where groups of structurally coherent elements are removed, conventional importance metrics frequently fail to maintain these essential performance attributes.<n>We propose an enhanced importance metric framework that not only reduces model size but also explicitly accounts for application-specific performance constraints.
  arXiv  Detail & Related papers  (2025-07-20T09:50:04Z)
• Comb, Prune, Distill: Towards Unified Pruning for Vision Model Compression [24.119415458653616]
  We propose a novel unified pruning framework Comb, Prune, Distill (CPD) to address both model-agnostic and task-agnostic concerns simultaneously. Our framework employs a combing step to resolve hierarchical layer-wise dependency issues, enabling architecture independence. In image classification we achieve a speedup of up to x4.3 with a accuracy loss of 1.8% and in semantic segmentation up to x1.89 with a 5.1% loss in mIoU.
  arXiv  Detail & Related papers  (2024-08-06T09:02:31Z)
• Iterative Soft Shrinkage Learning for Efficient Image Super-Resolution [91.3781512926942]
  Image super-resolution (SR) has witnessed extensive neural network designs from CNN to transformer architectures. This work investigates the potential of network pruning for super-resolution iteration to take advantage of off-the-shelf network designs and reduce the underlying computational overhead. We propose a novel Iterative Soft Shrinkage-Percentage (ISS-P) method by optimizing the sparse structure of a randomly network at each and tweaking unimportant weights with a small amount proportional to the magnitude scale on-the-fly.
  arXiv  Detail & Related papers  (2023-03-16T21:06:13Z)
• PAMS: Quantized Super-Resolution via Parameterized Max Scale [84.55675222525608]
  Deep convolutional neural networks (DCNNs) have shown dominant performance in the task of super-resolution (SR) We propose a new quantization scheme termed PArameterized Max Scale (PAMS), which applies the trainable truncated parameter to explore the upper bound of the quantization range adaptively. Experiments demonstrate that the proposed PAMS scheme can well compress and accelerate the existing SR models such as EDSR and RDN.
  arXiv  Detail & Related papers  (2020-11-09T06:16:05Z)
• DAIS: Automatic Channel Pruning via Differentiable Annealing Indicator Search [55.164053971213576]
  convolutional neural network has achieved great success in fulfilling computer vision tasks despite large computation overhead. Structured (channel) pruning is usually applied to reduce the model redundancy while preserving the network structure. Existing structured pruning methods require hand-crafted rules which may lead to tremendous pruning space.
  arXiv  Detail & Related papers  (2020-11-04T07:43:01Z)

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.