# Learning Source Acquisition Policies by Offline Planning

arXiv: https://arxiv.org/abs/2609.14299 · HTML: https://mlsanity.com/paper/2609.14299/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 13 of its 20 references that are in the map.
- First posted on arXiv 13 Sep 2026.

## Builds on (most cited first, 13 of 13)

- [Policy Distillation](https://mlsanity.com/paper/1511.06295/) (2015)
- [EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE](https://mlsanity.com/paper/1809.11142/) (2018)
- [Learning to Maximize Mutual Information for Dynamic Feature Selection](https://mlsanity.com/paper/2301.00557/) (2023)
- [Active Feature Acquisition with Generative Surrogate Models](https://mlsanity.com/paper/2010.02433/) (2020)
- [Learning to Search Better Than Your Teacher](https://mlsanity.com/paper/1502.02206/) (2015)
- [Classification with Costly Features as a Sequential Decision-Making Problem](https://mlsanity.com/paper/1909.02564/) (2019)
- [Estimating Conditional Mutual Information for Dynamic Feature Selection](https://mlsanity.com/paper/2306.03301/) (2023)
- [Acquisition Conditioned Oracle for Nongreedy Active Feature Acquisition](https://mlsanity.com/paper/2302.13960/) (2023)
- [Stochastic Encodings for Active Feature Acquisition](https://mlsanity.com/paper/2508.01957/) (2025)
- [AFABench: A Generic Framework for Benchmarking Active Feature Acquisition](https://mlsanity.com/paper/2508.14734/) (2025)
- [Sequential Cost-Sensitive Feature Acquisition](https://mlsanity.com/paper/1607.03691/) (2016)
- [Managing ML Pipelines: Feature Stores and the Coming Wave of Embedding Ecosystems](https://mlsanity.com/paper/2108.05053/) (2021)
- [Information Templates: A New Paradigm for Intelligent Active Feature Acquisition](https://mlsanity.com/paper/2508.18380/) (2025)

## Abstract (arXiv)

Predicting under an acquisition budget requires choosing feature groups whose value can depend on later queries. O-MPAC transfers finite-horizon risk-cost targets from complete training records into a shared source-action scorer. At inference time, the scorer uses partial observations and source metadata, re-scores after each query, and applies a hard cost mask. We analyze how tied teacher targets and the remaining planning horizon affect the learned decisions. Uniform supervision over tied minima preserves the target distribution under source relabeling. In a five-seed routing experiment, it achieves 0.965 accuracy under both original and context-last orders. On six real tasks, validation selects H1 without action cross-entropy in all thirty splits. O-MPAC has the highest mean budget-integrated accuracy on five tasks against source-adapted GDFS, DIME, AACO+NN and a static policy.