mlsanityData to 7 Oct 2026

Changing Model Behavior at Test-Time Using Reinforcement Learning

arXiv 1702.07780 ↗First posted 24 Feb 2017stat.ML cs.LGcited by 15

Not on the map. Why?

Abstract (arXiv)

Machine learning models are often used at test-time subject to constraints and trade-offs not present at training-time. For example, a computer vision model operating on an embedded device may need to perform real-time inference, or a translation model operating on a cell phone may wish to bound its average compute time in order to be power-efficient. In this work we describe a mixture-of-experts model and show how to change its test-time resource-usage on a per-input basis using reinforcement learning. We test our method on a small MNIST-based example.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). First posted on arXiv 24 Feb 2017.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 3; 2022: 2; 2023: 3; 2024: 3; 2025: 1; 2026: 3.

15 papers.

Sort:

By area:

  1. 2021Dynamic Neural Networks: A Surveycited by 405
  2. 2021AdaViT: Adaptive Tokens for Efficient Vision Transformercited by 360
  3. 2021Pruning and Quantization for Deep Neural Network Acceleration: A Surveycited by 259
  4. 2022Controllable Dynamic Multi-Task Architecturescited by 21
  5. 2022Task-Customized Self-Supervised Pre-training with Scalable Dynamic Routingcited by 18
  6. 2023ScrollNet: Dynamic Weight Importance for Continual Learningcited by 4
  7. 2023FIANCEE: Faster Inference of Adversarial Networks via Conditional Early Exitscited by 3
  8. 2025Energy-Aware Deep Learning on Resource-Constrained Hardwarecited by 3
  9. 2023Temporal Patience: Efficient Adaptive Deep Learning for Embedded Radar Data Processingcited by 2
  10. 2024Temporal Decisions: Leveraging Temporal Correlation for Efficient Decisions in Early Exit Neural Networkscited by 1
  11. 2024Predicting Probabilities of Error to Combine Quantization and Early Exiting: QuEEcited by 1
  12. 2026TTQ: Activation-Aware Test-Time Quantization to Accelerate LLM Inference On The Flycited by 1
3 more
  1. 2024Efficient Post-Training Augmentation for Adaptive Inference in Heterogeneous and Distributed IoT Environmentscited by 0
  2. 2026HERCULES: Hardware-Efficient, Robust, Continual Learning Neural Architecture Searchcited by 0
  3. 2026Exploring Budgeted Image Classification with Content-Sensitive Resource Allocationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026A Survey on Self-Improving Test-Time Intelligence: Feedback-Driven Adapting, Learning, and Scaling at Inference
  2. 2026Temporally Extended Mixture-of-Experts Models
  3. 2020Contextual Policy Transfer in Reinforcement Learning Domains via Deep Mixtures-of-Experts
  4. 2026On-the-Fly VLA Adaptation via Test-Time Reinforcement Learning
  5. 2025Scaling Test-Time Compute Without Verification or RL is Suboptimal
  6. 2026TTA-Vid: Generalized Test-Time Adaptation for Video Reasoning
  7. 2026Harnessing Image Question Dependence for Better VLM Test-time Reinforcement Learning
  8. 2025Local Mixtures of Experts: Essentially Free Test-Time Training via Model Merging