Early-Exit Neural Networks

Ask about this topic

Overview

Early-exit neural networks let each input use only as much model depth as it needs, aiming to save computation, time, and energy. Common approaches attach prediction heads to intermediate layers and use confidence thresholds or learned routers to decide whether to stop; training methods such as layer-wise supervision and distillation help make shallow predictions useful. The approach is studied in image classification and detection, speech, language-model reasoning, and edge-device inference, with evaluations measuring accuracy or task quality alongside latency, compute, and energy. Its benefits depend strongly on the model and workload: exits can hurt answer quality or add routing overhead, and confidence or output similarity does not always guarantee task correctness.

Latest papers

CardsList

44 papers3 papers in the last four weeks, down 67% on the four weeks before. 0.0% of all new papers.

  1. Client-Aware Security Adaptive Early Exit Vision Transformers for Heterogeneous Federated Learning

    Oct 16, 2026Sahar Refaei Ghalibafan, Samira Nazari, Christian Herglotz +2Early-Exit Neural NetworksHeterogeneous Federated Learning

    Brandenburg University of Technology, Cottbus, Germany · University of Zanjan, Zanjan, Iran · Tallinn University of Technology, Tallinn, Estonia +1

  2. Selection-Safe Conformal Early Exit: Distribution-Free Coverage under Adaptive Depth

    Oct 14, 2026Adria Binte Habib, Ibne Farabi ShihabConformal PredictionAdaptive Inference

    Department of Computer Science, BRAC University · Department of Computer Science, Iowa State University

  3. DEX: Digit-Level Early Exit for Energy-Efficient MSDF Neural Network Inference

    Oct 8, 2026Yousef Sadegheih, Dorit Merhof, Muhammad UsmanEfficient Neural Network InferenceEarly-Exit Neural Networks

    Faculty of Informatics and Data Science, University of Regensburg, 93053 Regensburg, Germany

  4. When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection

    Sep 26, 2026Tianyang Zhou, Leman AkogluEarly-Exit Neural NetworksAnomaly Detection

    Carnegie Mellon University

  5. Does per-frame early exit pay? A compute-matched study of dynamic depth for on-device speech enhancement

    Sep 24, 2026Clément Laroche, Riccardo MicciniEfficient Neural Network InferenceSpeech Enhancement

    GN A/S, Denmark · Technical University of Denmark, Denmark

  6. Traffic Sign Recognition for Autonomous Driving Using Branched YOLOv2 and Geometric Features

    Sep 18, 2026Arefeh RezaeiTraffic Sign DetectionAutonomous Driving Perception

    Department of Computer Engineering, K. N. Toosi University of Technology

  7. Proportional-Fair Resource Allocation and Dual-Threshold Early-Exit Inference for Secure Cooperative Multi-Layer Edge Intelligence

    Sep 14, 2026Thai T. Vu, John Le, Tu N. Nguyen +3Edge InferenceFair Resource Allocation

    School of Computing and Information Technology, University of Wollongong, Wollongong, NSW, Australia · Department of Computer Science, Kennesaw State University, Georgia, USA · CodeZX Software Company Limited, Vietnam

  8. Adaptive AI: Energy Efficient Multi-exit TinyML on Intelligent Vision Systems at the Edge

    Sep 14, 2026Luca Crupi, Lorenzo Lamberti, Alessandro Giusti +1Efficient Neural Network InferenceEnergy-Efficient ML

    IDSIA, SUPSI, Lugano, Switzerland · IIS, ETH Zürich, Switzerland

  9. One Size Does Not Fit All: Setting Inference Depth from the Questions a Deployment Actually Asks

    Sep 12, 2026Jerry KaplanEfficient Language Model InferenceLLM Inference Acceleration

    Adjunct Lecturer, Computer Science and Masters in International Policy, Stanford University

  10. Theoretical Guarantees for One-Shot Magnitude Pruning and Compute-Adaptive Early Exit

    Sep 11, 2026Erdem KoyuncuNeural Network GeneralizationEarly-Exit Neural Networks

    Department of Electrical and Computer Engineering, University of Illinois Chicago, IL, USA.

  11. Adaptive Gated Deepfake Detection for Low-Resolution and Resource-Constrained Environments

    Sep 4, 2026Vaishnavi Sen, Cody Laurie, Rashida HasanDeepfake DetectionAdaptive Inference

    Department of Computer Science California State University, Northridge California, USA

  12. </think> Doesn't Stop Reasoning: Analysis of Spurious CoT Termination

    Sep 3, 2026Seunghee Koh, Sungjae Choi, Minchan Kwon +2CoT ReasoningEfficient Language Model Reasoning

    Korea Advanced Institute of Science and Technology, South Korea

  13. ENAF: A Multi-Exit Network with an Adaptive Patch Fusion for Large Image Super Resolution

    Aug 15, 2026Duong M. Nguyen, Tuan Nghia Nguyen, Xuan Truong NguyenImage Super-ResolutionEfficient Neural Network Inference

    Hanoi University of Science and Technology Hanoi, Vietnam · Seoul National University South Korea, Seoul 08826

  14. Adaptive-WAM: Quality-Guided Early-Exit Planning from Intermediate Video-Diffusion Features

    Aug 6, 2026Sining Ang, Yuguang Yang, Yan WangEfficient InferenceDiffusion Transformer

    Institute for AI Industry Research (AIR), Tsinghua University · Department of Automation, University of Science and Technology of China · School of Electronic Information Engineering, Beihang University

  15. Diversity is Not Ambiguity: Toward Accurate and Efficient Ambiguity Detection for Open-Domain QA

    Aug 4, 2026Jiwon Lee, Yong-chan Park, Jungin Hong +1Question AnsweringOpen-Domain QA

    Seoul National University

  16. Stop When Memory Suffices: Evidence-Conditioned Progressive Execution for LLM Agents

    Aug 2, 2026Yidan Lin, Kaixiang Wang, Jiong Lou +1LLM Inference EfficiencyEfficient Inference

  17. Real-Time Visual Obstruction Detection in Surgical Augmented Reality

    Jul 31, 2026Shih-Chin Yang, Yanming Xiu, Hanting Ye +3Efficient InferenceMedical VLMs

    Department of Computer Science, Duke University · Department of Electrical and Computer Engineering, Duke University

  18. BLADE: Boundary-Expanded and Layer-Adaptive Dynamic Exit for Efficient LLM Reasoning

    Jul 31, 2026Keshu Fu, Keqin Peng, Jun Bai +6Efficient Language Model ReasoningEfficient Language Model Inference

    School of Computer Science and Engineering, Beihang University · Beijing Institute for General Artificial Intelligence, BIGAI · Academy for Advanced Interdisciplinary Studies, Peking University +1

  19. Fast Wireless Foundation Models with Early-Exits

    Jun 28, 2026Omar Mashaal, Hatem Abou-ZeidWireless Foundation ModelsEfficient Inference

    Department of Electrical and Software Engineering, University of Calgary, Calgary, AB T2N 1N4, Canada

  20. Early-Exit Graph Neural Networks for Link Prediction

    Jun 20, 2026Roman Knyazhitskiy, Andrea Giuseppe Di FrancescoEfficient Neural Network InferenceGraph Neural Networks

    University of Cambridge · Sapienza University of Rome

  21. Think Less, Act Early: Reinforced Latent Reasoning with Early Exit in Vision-Language-Action Models

    Jun 13, 2026Dianqiao Lei, Lianlei ShanEfficient InferenceEfficient VLA Model Inference

    1Tsinghua University, Beijing, China.

  22. ViT-FREE: Efficient Face Recognition via Early Exiting and Synthetic Adaptation

    Jun 10, 2026Tahar Chettaoui, Guray Ozgur, Eduarda Caldeira +2Inference-Time OptimizationEfficient ViTs

    Fraunhofer Institute for Computer Graphics Research IGD, Germany · Department of Computer Science, TU Darmstadt, Germany

  23. TabSwift: An Efficient Tabular Foundation Model with Row-Wise Attention

    Jun 5, 2026Si-Yang Liu, Han-Jia YeTabular Foundation ModelsEfficient Inference

    School of Artificial Intelligence, Nanjing University, China · National Key Laboratory for Novel Software Technology, Nanjing University, China.

  24. t-gems: text-guided exit modules for decreasing clip image encoder

    May 17, 2026Alberto Presta, Grzegorz Stefanski, Michal Byra +1Cross-Modal LearningVision-Language Models

    Samsung AI Center Warsaw · Institute of Fundamental Technological Research, PAS, Warsaw

  25. Trajectory-Aware Adaptive Inference in Object Detection Models

    May 12, 2026Grigorios Papanikolaou, Ioannis Kontopoulos, Giannis Spiliopoulos +2Adaptive InferenceDynamic Neural Networks

    Department of Electrical and Computer Engineering, National Technical University of Athens, Greece · Department of Product and Systems Design Engineering, University of the Aegean, Syros, Greece

  26. Amortized-Precision Quantization for Early-Exit Vision Transformers

    May 8, 2026Rui Fang, Hsi-Wen Chen, Ming-Syan ChenMixed-Precision QuantizationDynamic Neural Networks

    National Taiwan University

  27. Balancing Stability and Plasticity in Sequentially Trained Early-Exiting Neural Networks

    May 6, 2026Alaa Zniber, Ouassim Karrakchou, Mounir GhoghoAdaptive InferenceDynamic Neural Networks

    ⋆TICLab, International University of Rabat, Morocco · ‡ School of Electronic and Electrical Engineering, University of Leeds, UK

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

    May 1, 2026Shashank Kapadia, Deep Naryan Mishra, Sujal Reddy Alugubelli +4Efficient Transformer InferenceDynamic Neural Networks

    Walmart Inc. · Sam’s Club

  29. A Comparative Analysis on the Performance of Upper Confidence Bound Algorithms in Adaptive Deep Neural Networks

    Apr 27, 2026Grigorios Papanikolaou, Ioannis Kontopoulos, Konstantinos TserpesMulti-Armed BanditsDynamic Neural Networks

    Department of Electrical and Computer Engineering National Technical University of Athens, Greece

  30. Progressive Approximation in Deep Residual Networks: Theory and Validation

    Apr 27, 2026Wei Wang, Xiao-Yong Wei, Qing LiEfficient Neural Network InferenceResidual Learning

    Department of Computing, the Hong Kong Polytechnic University.