mlsanityData to 7 Oct 2026

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

arXiv 2609.12337 ↗First posted 11 Sep 2026cs.LGcited by 0

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

Abstract (arXiv)

We study compute reduction in neural networks through a unified partial versus full computation view, captured by one-shot magnitude pruning in the static regime and early exit in the adaptive regime. In an asymptotic single-neuron model, we prove a concentration theorem for one-shot magnitude pruning with explicit rates. We also introduce the conditional perceptron for early exit and show that its excess generalization error decays as a power of the compute gap, with an exponent that grows to infinity as the alignment between partial and full computations tends to one. We then extend the analysis to deep networks, characterizing how pruning-induced distortions accumulate with depth and deriving a corresponding compute-accuracy tradeoff for frozen-backbone early exit under a neural network Gaussian process model. Numerical simulations corroborate the predicted scaling laws.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Early exit and edge inference, one of 12 directions in Efficient LLM inference and serving. Builds on 39 of its 70 references that are in the map. First posted on arXiv 11 Sep 2026.

Builds on

39 of 70 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025HELIOS: Adaptive Model And Early-Exit Selection for Efficient LLM Inference Serving14 Apr
  2. 2025Wanda++: Pruning Large Language Models via Regional Gradients6 Mar
  3. 2024Fast yet Safe: Early-Exiting with Risk Control31 May
  4. 2021Adaptive Inference through Early-Exit Networks: Design, Challenges and Directions9 Jun

Other references · 35

  1. 2020Scaling Laws for Neural Language Modelscited by 6,411
  2. 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,413
  3. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
  4. 2015Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Codingcited by 2,291
  5. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  6. 2024Mixtral of Expertscited by 1,785
  7. 2023SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shotcited by 1,090
  8. 2023A Simple and Effective Pruning Approach for Large Language Modelscited by 776
  9. 2017BranchyNet: Fast Inference via Early Exiting from Deep Neural Networkscited by 591
  10. 2022Beyond neural scaling laws: beating power law scaling via data pruningcited by 519
  11. 2021Dynamic Neural Networks: A Surveycited by 405
  12. 2018Gaussian Process Behaviour in Wide Deep Neural Networkscited by 314
23 more
  1. 2013Learning Factored Representations in a Deep Mixture of Expertscited by 275
  2. 2024Scaling Laws for Fine-Grained Mixture of Expertscited by 166
  3. 2019Tensor Programs I: Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processescited by 148
  4. 2016Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivitycited by 138
  5. 2018Deep Neural Decision Treescited by 72
  6. 2024MaskLLM: Learnable Semi-Structured Sparsity for Large Language Modelscited by 49
  7. 2013Low-Rank Approximations for Conditional Feedforward Computation in Deep Neural Networkscited by 44
  8. 2018Data-Dependent Coresets for Compressing Neural Networks with Applications to Generalization Boundscited by 43
  9. 2024Everybody Prune Now: Structured Pruning of LLMs with only Forward Passescited by 42
  10. 2023SEENN: Towards Temporal Spiking Early-Exit Neural Networkscited by 39
  11. 2019Data-Independent Neural Pruning via Coresetscited by 35
  12. 2016Neural Random Forestscited by 25
  13. 2025Adaptive Layer-skipping in Pre-trained LLMscited by 24
  14. 2023Unified Data-Free Compression: Pruning and Quantization without Fine-Tuningcited by 22
  15. 2022Pruning Neural Networks via Coresets and Convex Geometry: Towards No Assumptionscited by 20
  16. 2019SiPPing Neural Networks: Sensitivity-informed Provable Pruning of Neural Networkscited by 17
  17. 2022Unsupervised Early Exit in DNNs with Multiple Exitscited by 10
  18. 2023Data-Free Quantization via Mixed-Precision Compensation without Fine-Tuningcited by 9
  19. 2021E2CM: Early Exit via Class Means for Efficient Supervised and Unsupervised Learningcited by 8
  20. 2023Early-Exit Neural Networks with Nested Prediction Setscited by 6
  21. 2022Class Based Thresholding in Early Exit Semantic Segmentation Networkscited by 5
  22. 2023Memorization Capacity of Neural Networks with Conditional Computationcited by 3
  23. 2025Training-Free Restoration of Pruned Neural Networkscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2022Pruning Early Exit Networks
  2. 2021Only Train Once: A One-Shot Neural Network Training And Pruning Framework
  3. 2019Structural Pruning in Deep Neural Networks: A Small-World Approach
  4. 2025FAIR-Pruner: A Flexible Framework for Automatic Layer-Wise Pruning via Tolerance of Difference
  5. 2023PERP: Rethinking the Prune-Retrain Paradigm in the Era of LLMs
  6. 2026Exploring the Limits of Pruning: Task-Specific Neurons, Model Collapse, and Recovery in Task-Specific Large Language Models
  7. 2026A Comparative Study of CNN Optimization Methods for Edge AI: Exploring the Role of Early Exits
  8. 2024Is Oracle Pruning the True Oracle?