# Transfer Calibrated Prediction Powered Inference

arXiv: https://arxiv.org/abs/2609.34156 · HTML: https://mlsanity.com/paper/2609.34156/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 44 references that are in the map.
- First posted on arXiv 28 Sep 2026.

## Builds on (most cited first, 8 of 8)

- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [PPI++: Efficient Prediction-Powered Inference](https://mlsanity.com/paper/2311.01453/) (2023)
- [AutoEval Done Right: Using Synthetic Data for Model Evaluation](https://mlsanity.com/paper/2403.07008/) (2024)
- [RITA: a Study on Scaling Up Generative Protein Sequence Models](https://mlsanity.com/paper/2205.05789/) (2022)
- [Heterogeneous transfer learning for high-dimensional regression with feature mismatch](https://mlsanity.com/paper/2412.18081/) (2024)
- [Prediction-powered Inference by Mixture of Experts](https://mlsanity.com/paper/2604.27892/) (2026)
- [Recidivism Prediction, Peer Effect Estimation, and Prediction-Powered Inference with LLM Text Measures](https://mlsanity.com/paper/2509.20634/) (2025)

## Abstract (arXiv)

Prediction-powered inference (PPI) and its power-tuned extension (PPI++) improve confidence intervals by combining a small gold-standard labeled sample with a large AI model's predictions. Its efficiency gain relies on low residual variance, which may not hold if the predictor is pre-trained on a different source domain. We propose Transfer Calibrated Prediction-Powered Inference (TC-PPI), adapting the source-domain predictor to the target domain using gold-standard samples through cross-fitting. This approach supports various adaptation methods, such as sparse linear calibration, LoRA, and fine-tuning. Our jointly tuned cross-fit estimator, Joint-TC-Cross-PPI++, maintains unbiasedness and is simultaneously at least as efficient as classical inference, PPI, and PPI++, thereby protecting against negative transfer. We provide high-dimensional MSE bounds for calibration and show empirical improvements over baseline methods across various real-world applications.