LEAP: Layer-wise Exit-Aware Pretraining for Efficient Transformer Inference

• URL: http://arxiv.org/abs/2605.01058v1
• Date: Fri, 01 May 2026 19:45:26 GMT
• Title: LEAP: Layer-wise Exit-Aware Pretraining for Efficient Transformer Inference
• Abstract summary: Layer-aligned distillation and convergence-based early exit represent two predominant computational efficiency paradigms for transformer inference.<n>We introduce LEAP (Layer-wise Exit-Aware Pretraining), an auxiliary training objective that reconciles this incompatibility.<n>LEAP-MiniLM achieves 1.61$times$ measured wall-clock speedup, with 91.9% of samples exiting by layer 7 and 1.80$times$ theoretical layer reduction.
• Score: 0.5277024349608834
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Layer-aligned distillation and convergence-based early exit represent two predominant computational efficiency paradigms for transformer inference; yet we establish that they exhibit systematic incompatibility under standard deployment conditions for convergence-based early exit. Distillation objectives that align intermediate student layers to teacher representations suppress the representational convergence that early-exit mechanisms exploit, rendering such mechanisms ineffective on distilled models. We introduce LEAP (Layer-wise Exit-Aware Pretraining), an auxiliary training objective that reconciles this incompatibility. LEAP requires no architectural modifications; it augments standard distillation with a single constraint ensuring intermediate layers approximate final-layer representations. LEAP-MiniLM achieves 1.61$\times$ measured wall-clock speedup (batch=1, NVIDIA L4) at $θ$=0.95, with 91.9% of samples exiting by layer 7 and 1.80$\times$ theoretical layer reduction, where standard distilled models achieve zero effective speedup. We validate across sentence similarity (STS-B: 0.760 $\pm$ 0.006) and retrieval benchmarks (BEIR), providing operational guidance including latency measurements, decision thresholds, and deployment criteria.

Related papers

• Self-Adversarial One Step Generation via Condition Shifting [11.426065945265647]
  APEX is a framework for adversarial correction signals extracted endogenously from a flow model.<n>Our 0.6B model surpasses FLUX-Schnell 12B (20$times$ more parameters) in one step quality.<n>With LoRA tuning on Qwen-Image 20B, APEX reaches a GenEval score of 0.89 at NFE=1 in 6 hours, surpassing the original 50-step teacher (0.87) and providing a 15.33$times$ inference speedup.
  arXiv  Detail & Related papers  (2026-04-14T05:54:33Z)
• A1: A Fully Transparent Open-Source, Adaptive and Efficient Truncated Vision-Language-Action Model [112.9420001646428]
  Vision-Language-Action (VLA) models have emerged as a powerful paradigm for open-world robot manipulation, but their practical deployment is often constrained by cost.<n>We present A1, a fully open-source and transparent VLA framework designed for low-cost, high- throughput inference.<n>A1 achieves state-of-the-art success rates while significantly reducing inference cost.
  arXiv  Detail & Related papers  (2026-04-07T10:18:40Z)
• SIMPLER: Efficient Foundation Model Adaptation via Similarity-Guided Layer Pruning for Earth Observation [3.1313965020437546]
  We introduce SIMPLER, a pre-fine-tuning architecture selection method that reduces inference and deployment costs.<n>On Prithvi-EO-2, SIMPLER prunes up to 79% of parameters while retaining 94% of baseline performance, yielding a 2.1x training speedup and 2.6x inference speedup.
  arXiv  Detail & Related papers  (2026-03-20T11:38:32Z)
• EcoSpa: Efficient Transformer Training with Coupled Sparsity [79.5008521101473]
  Transformers have become the backbone of modern AI, yet their high computational demands pose critical system challenges.<n>We introduce EcoSpa, an efficient structured sparse training method that jointly evaluates and sparsifies coupled weight matrix pairs.
  arXiv  Detail & Related papers  (2025-11-09T11:23:43Z)
• Distilling Parallel Gradients for Fast ODE Solvers of Diffusion Models [53.087070073434845]
  Diffusion models (DMs) have achieved state-of-the-art generative performance but suffer from high sampling latency due to their sequential denoising nature.<n>Existing solver-based acceleration methods often face image quality degradation under a low-latency budget.<n>We propose the Ensemble Parallel Direction solver (dubbed as ours), a novel ODE solver that mitigates truncation errors by incorporating multiple parallel gradient evaluations in each ODE step.
  arXiv  Detail & Related papers  (2025-07-20T03:08:06Z)
• Determining Layer-wise Sparsity for Large Language Models Through a Theoretical Perspective [55.90119819642064]
  We address the challenge of determining the layer-wise sparsity rates of large language models (LLMs) through a theoretical perspective.<n>This refers to the cumulative effect of reconstruction errors throughout the sparsification process.<n>We derive a simple yet effective approach to layer-wise sparsity allocation that mitigates this issue.
  arXiv  Detail & Related papers  (2025-02-20T17:51:10Z)
• Token Compensator: Altering Inference Cost of Vision Transformer without Re-Tuning [63.43972993473501]
  Token compression expedites the training and inference of Vision Transformers (ViTs) However, when applied to downstream tasks, compression degrees are mismatched between training and inference stages. We propose a model arithmetic framework to decouple the compression degrees between the two stages.
  arXiv  Detail & Related papers  (2024-08-13T10:36:43Z)
• Inverse-Free Fast Natural Gradient Descent Method for Deep Learning [52.0693420699086]
  We present a fast natural gradient descent (FNGD) method that only requires inversion during the first epoch. FNGD exhibits similarities to the average sum in first-order methods, leading to the computational complexity of FNGD being comparable to that of first-order methods.
  arXiv  Detail & Related papers  (2024-03-06T05:13:28Z)
• Constraint-aware and Ranking-distilled Token Pruning for Efficient Transformer Inference [18.308180927492643]
  ToP is a ranking-distilled token distillation technique, which distills effective token rankings from the final layer of unpruned models to early layers of pruned models. ToP reduces the average FLOPs of BERT by 8.1x while achieving competitive accuracy on GLUE, and provides a real latency speedup of up to 7.4x on an Intel CPU.
  arXiv  Detail & Related papers  (2023-06-26T03:06:57Z)
• COST-EFF: Collaborative Optimization of Spatial and Temporal Efficiency with Slenderized Multi-exit Language Models [16.586312156966635]
  Transformer-based pre-trained language models (PLMs) mostly suffer from excessive overhead despite their advanced capacity. Existing statically compressed models are unaware of the diverse complexities between input instances. We propose a collaborative optimization for PLMs that integrates static model compression and dynamic inference acceleration.
  arXiv  Detail & Related papers  (2022-10-27T15:06:40Z)
• Layer Reduction: Accelerating Conformer-Based Self-Supervised Model via Layer Consistency [31.572652956170252]
  Transformer-based self-supervised models are trained as feature extractors and have empowered many downstream speech tasks to achieve state-of-the-art performance. We experimentally achieve 7.8X parameter reduction, 41.9% training speedup and 37.7% inference speedup while maintaining comparable performance with conventional BERT-like self-supervised methods.
  arXiv  Detail & Related papers  (2021-04-08T08:21:59Z)

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.