mlsanityData to 14 Oct 2026

CalexNet: Soft Cascade-Aligned Training and Calibration for Lightweight Early-Exit Branches

arXiv 2509.08318 ↗First posted 10 Sep 2025cs.CVcited by 0

On the map: Efficient LLM inference and serving → Early exit and edge inference

Abstract (arXiv)

Early-exit cascades over a frozen convolutional backbone enable adaptive inference but suffer from three sources of train-inference mismatch: branches train on samples they will never see at inference, their per-class precision thresholds are calibrated on the wrong distribution, and the standard cross-entropy target on backbone argmax labels discards the backbone's uncertainty signal. We close all three gaps with CalexNet (Cascade-Aligned Early eXits), a training-recipe-only modification: branches train under continuously-weighted importance sampling that matches the cascade-survivor distribution; per-class precision thresholds are calibrated on the actual cascade-survivor subset of the validation set; and the classification head is trained against the backbone's full softmax via a temperature-scaled KL objective. Combined with an augmented prototype-pooling branch head, CalexNet is evaluated on ResNet18 and ResNet50 backbones across CIFAR-100 (20-superclass coarse, the harder primary setting) and CINIC-10 (10-class, the easier cross-validation counterpart). On the accuracy-FLOPs Pareto frontier, CalexNet matches or exceeds three published baselines (PTEEnet, ZTW, BoostNet) and a within-paper "no-alignment, no-KD" reference. The largest gains appear in the practically relevant 30-70% FLOPs-reduction regime and are stable across n=3 training seeds. CalexNet requires no inference-time architectural change and is a drop-in for any frozen-backbone early-exit cascade.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Early exit and edge inference, one of 12 directions in Efficient LLM inference and serving. Builds on 24 of its 40 references that are in the map. First posted on arXiv 10 Sep 2025.

Builds on

24 of 40 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025DistrEE: Distributed Early Exit of Deep Neural Network Inference on Edge Devices6 Feb
  2. 2025PTEENet: Post-Trained Early-Exit Neural Networks Augmentation for Inference Cost Optimization5 Jan
  3. 2024LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding25 Apr
  4. 2022Predictive Exit: Prediction of Fine-Grained Early Exits for Computation- and Energy-Efficient Inference9 Jun
  5. 2021Adaptive Inference through Early-Exit Networks: Design, Challenges and Directions9 Jun
  6. 2021Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges8 Mar

Other references · 18

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  3. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 2,031
  6. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  7. 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,593
  8. 2017Cascade R-CNN: Delving into High Quality Object Detectioncited by 1,276
  9. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,245
  10. 2017BranchyNet: Fast Inference via Early Exiting from Deep Neural Networkscited by 596
  11. 2021Dynamic Neural Networks: A Surveycited by 408
  12. 2021Pruning and Quantization for Deep Neural Network Acceleration: A Surveycited by 263
6 more
  1. 2018What is the Effect of Importance Weighting in Deep Learning?cited by 251
  2. 2023A Survey on Deep Neural Network Pruning-Taxonomy, Comparison, Analysis, and Recommendationscited by 202
  3. 2019Improved Techniques for Training Adaptive Deep Networkscited by 66
  4. 2023Jointly-Learned Exit and Inference for a Dynamic Neural Network : JEI-DNNcited by 15
  5. 2023Computation-efficient Deep Learning for Computer Vision: A Surveycited by 14
  6. 2022Boosted Dynamic Neural Networkscited by 8

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023Window-Based Early-Exit Cascades for Uncertainty Estimation: When Deep Ensembles are More Efficient than Single Models
  2. 2026SAFE-KD: Risk-Controlled Early-Exit Distillation for Vision Backbones
  3. 2026Predicting Inference-Time Scaling Gains from Labeled Validation-Set Output Statistics
  4. 2025Confidence-gated training for efficient early-exit neural networks
  5. 2026Adaptive Layerwise Perturbation: Unifying Off-Policy Corrections for LLM RL
  6. 2026Balancing Stability and Plasticity in Sequentially Trained Early-Exiting Neural Networks
  7. 2023Finding the SWEET Spot: Analysis and Improvement of Adaptive Inference in Low Resource Settings
  8. 2026Variational Filtering for Bayesian Last Layers with an Independent Binary Probit Surrogate