AI Model Modulation with Logits Redistribution
Not on the map. Why?
Abstract (arXiv)
Large-scale models are typically adapted to meet the diverse requirements of model owners and users. However, maintaining multiple specialized versions of the model is inefficient. In response, we propose AIM, a novel model modulation paradigm that enables a single model to exhibit diverse behaviors to meet the specific end requirements. AIM enables two key modulation modes: utility and focus modulations. The former provides model owners with dynamic control over output quality to deliver varying utility levels, and the latter offers users precise control to shift model's focused input features. AIM introduces a logits redistribution strategy that operates in a training data-agnostic and retraining-free manner. We establish a formal foundation to ensure AIM's regulation capability, based on the statistical properties of logits ordering via joint probability distributions. Our evaluation confirms AIM's practicality and versatility for Al model modulation, with tasks spanning image classification, semantic segmentation and text generation, and prevalent architectures including ResNet, SegFormer and Llama.
Key facts. Cited by 3 papers in the map (data through 13 Oct 2026). Builds on 17 of its 53 references that are in the map. First posted on arXiv 13 Mar 2026.
Builds on
17 of 53 references.
References · 17
- 2017Attention Is All You Needcited by 60,106
- 2015Deep Residual Learning for Image Recognitioncited by 51,950
- 2020Language Models are Few-Shot Learnerscited by 32,320
- 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,072
- 2021Training Verifiers to Solve Math Word Problemscited by 9,916
- 2020Measuring Massive Multitask Language Understandingcited by 7,876
- 2017On Calibration of Modern Neural Networkscited by 4,902
- 2014Fully Convolutional Networks for Semantic Segmentationcited by 4,849
- 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,388
- 2022Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learningcited by 891
- 2017BranchyNet: Fast Inference via Early Exiting from Deep Neural Networkscited by 592
- 2020BERT Loses Patience: Fast and Robust Inference with Early Exitcited by 270
5 more
- 2023When Large Language Models Meet Personalization: Perspectives of Challenges and Opportunitiescited by 90
- 2022Predictive Exit: Prediction of Fine-Grained Early Exits for Computation- and Energy-Efficient Inferencecited by 24
- 2023Window-Based Early-Exit Cascades for Uncertainty Estimation: When Deep Ensembles are More Efficient than Single Modelscited by 15
- 2024Exploring ChatGPT App Ecosystem: Distribution, Deployment and Securitycited by 12
- 2025Re-Key-Free, Risky-Free: Adaptable Model Usage Controlcited by 6
Built on by
Citing papers per year: 2026: 3.
3 papers.