The paper introduces a new framework for activation steering in large language models, distinguishing between the steering direction (estimand) and the method used to recover it (estimator). It shows that most existing techniques can be viewed as plug‑in estimators of the average gradient of an outcome regression, and proposes a complementary estimator based on the Stein score to directly target the gradient. By combining these approaches, the authors achieve an orthogonalized estimator with double‑robustness properties and demonstrate its effectiveness on benchmark datasets.
arXiv:2411.02771v3 Announce Type: replace-cross Abstract: Doubly robust estimators are widely used for estimating average treatment effects and other linear summaries of regression functions. While c...
The paper systematically evaluates regression objectives for tabular deep learning across many datasets and neural backbones, finding that most alternatives to Mean Squared Error (MSE) offer limited or inconsistent gains, while quantile loss consistently improves performance. It identifies a limitation of global early stopping: different noise levels in target regions reach optimal performance at different epochs, causing a compromise. To address this, the authors propose a heteroskedasticity‑aware extension of quantile loss using input‑dependent label smoothing, which outperforms standard quantile loss on several backbones.
The paper introduces OTROPE, a likelihood‑free method for off‑policy evaluation of large language models (LLMs) that uses optimal transport to align labeled samples from a behavior model with unlabeled samples from a target model in a semantic space. OTROPE corrects human‑labeled residuals with proxy predictors, achieving a doubly robust evaluation without requiring behavior‑policy modeling or density‑ratio estimation. The authors provide theoretical guarantees for consistency and convergence, and demonstrate through synthetic and real LLM tasks that OTROPE outperforms existing baselines and can elevate weaker evaluators to match or exceed stronger ones.
arXiv:2606. 25188v1 Announce Type: new Abstract: Efficient uncertainty quantification (UQ) is essential for trustworthy large-scale learning.
arXiv:2607. 23388v1 Announce Type: cross Abstract: As constrained learning becomes increasingly common, models are trained under explicit feasibility requirements to enforce fairness, safety, robustness, regulariza- tion, and physics or logic constraints.
arXiv:2608. 18339v1 Announce Type: cross Abstract: Vision-language models (VLMs) have demonstrated remarkable zero-shot capabilities yet remain sensitive to real-world distribution shifts during inference.
arXiv:2606. 08454v1 Announce Type: new Abstract: Activation steering provides a lightweight inference-time mechanism for controlling large language models (LLMs) by modifying their internal activation vectors toward desired behaviors.
arXiv:2607. 08347v1 Announce Type: cross Abstract: Active testing provides a label--efficient approach to risk estimation by adaptively selecting which test points should be labelled.
The paper introduces BROT, a two‑step approach for estimating optimal transport maps. First, it computes the unregularized OT plan, then fits a deep neural network to the resulting barycentric targets using least‑squares regression. The authors prove that, under standard regularity conditions, BROT achieves the minimax convergence rate when the true OT map is Lipschitz, and demonstrate its effectiveness on synthetic data, images, and downstream tasks such as single‑cell perturbation prediction and unsupervised domain adaptation.
arXiv:2606. 05165v1 Announce Type: new Abstract: Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data.
arXiv:2610.07804v1 Announce Type: new Abstract: Prior Fitted Networks (PFNs) such as TabPFN now rival established statistical procedures across prediction and estimation tasks. A natural explanation...
The paper introduces the Directional Influence Function (DIF), a new estimator for measuring how training samples affect the solution of constrained learning problems. DIF incorporates constraints directly by formulating optimality conditions as a variational inequality, thereby avoiding the bias seen in classical influence functions. Experiments on constrained linear regression and fairness‑constrained CNNs show that DIF accurately predicts changes in test loss and aligns closely with actual retraining results.