Tag
This paper proposes CRUMB, a three-stage inference wrapper that clusters test queries and selects a distributionally matched training subset via MMD minimization to enable efficient Prior-Fitted Network inference on large datasets, achieving state-of-the-art context selection on 51 TabArena datasets.
Introduces CICL, a decision-aware context layer that selects and compresses evidence for tool-using LLM agents by treating context as a decision-time intervention, using counterfactual-inspired scoring and typed memory cards under a token budget. Experiments on SWE-bench and RepoBench show concrete gains in retrieval accuracy and action criticality.
Proposes G²C-MT, a graph-guided context selection framework for document-level machine translation that models structured discourse dependencies via a lightweight discourse graph and depth-biased random walk, outperforming baselines on multiple LLMs.