leakage

The paper introduces a leakage-aware framework for conversational task disambiguation over tabular data, decomposing agents into asking and solution policies, and presents AmbiTab, a benchmark suite unifying six ambiguous text-to-SQL datasets. It also trains an asking policy with reinforcement learning, improving disambiguation metrics on all six datasets and task success on five.

This paper empirically investigates shortcut learning in legal judgment prediction using UK Employment Tribunal claims, finding that models exploit outcome-revealing cues in post-hoc judicial texts, but performance remains robust after masking such leakage features.

MosaicLeaks introduces a new benchmark for measuring privacy leakage in deep-research AI agents, showing that agents often leak private information through external queries and proposing a training method (PA-DR) to reduce leakage while improving task performance.

Introduces MosaicLeaks, a benchmark of 1,001 multi-hop deep research tasks that chain private enterprise documents with public web queries to evaluate privacy leakage. Finds that models leak sensitive information at multiple levels, and proposes PA-DR, a reinforcement learning framework that reduces leakage while improving task accuracy.