Minimax PAC Bounds for Learning in Exogenous Contextual MDPs
2026-06-25T07:23:56Z•3dfb7d7750f4eb27fd93142c05d7b6ed8a88d4faa8109fa9d4cee6cd87cea57e
PAC-boundsPolyak-RuppertTD-learningblack-box-stabilityclass-imbalancecontextual-MDPdata-augmentationdiffusion-modelse-valuesfederated-learninghyperparameter-selectioninformation-theory','pac-bayes','ranking','partial-order-modelslarge-language-modelslearn-then-testmachine-learningmean-fieldp-valuesrandomizationreinforcement-learningsample-complexitystatisticssynthetic-augmentationtemporal-point-processesvariance-reductionvariational-inference
What happened
Collection of new arXiv submissions (2026-06-25) in statistics and machine learning covering theoretical and algorithmic advances: minimax PAC/sample-complexity bounds for tabular contextual MDPs with exogenous i.i.d. contexts (variance-reduced, |Z|-free rates where applicable); a task-oriented randomization framework for stabilizing black-box models (theory and LLM-motivated top-k extensions); a unified statistical learn-then-test (LTT) framework for hyperparameter selection with finite-sample guarantees using p-values/e-values; analysis showing Gaussian mean-field variational inference can (
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_stat_ml
- Record identifier
- 3dfb7d7750f4eb27fd93142c05d7b6ed8a88d4faa8109fa9d4cee6cd87cea57e
- Enrichment time
- 2026-06-25T07:23:56Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.