Search papers, labs, and topics across Lattice.
This paper introduces EdgeLM, a novel retrieval framework that enhances table-centric predictions in large language models (LLMs) by selecting edge evidence鈥攄emonstrations that are both relevant and informative about decision boundaries. Unlike traditional methods that focus on similarity, EdgeLM retrieves two types of edge evidence: data edges, which are nearby examples with differing labels, and model edges, which are similar examples previously misclassified by the model. The framework consistently outperforms existing methods across various data wrangling tasks and LLMs, demonstrating the effectiveness of its complementary evidence selection approach without requiring model retraining or task-specific adjustments.
EdgeLM reveals that selecting edge evidence can significantly improve LLM performance in table understanding tasks, outperforming traditional similarity-based retrieval methods.
Large language models (LLMs) perform table-centric prediction through in-context learning, making demonstration selection critical to performance. Existing retrieval methods prioritize similarity to the query, but similar demonstrations often reinforce the model's likely prediction rather than reveal the distinctions needed for difficult decisions. We propose EdgeLM, a retrieval framework that instead selects edge evidence, demonstrations that are both relevant to the query and informative about the decision boundary. EdgeLM retrieves two complementary forms of edge evidence by selecting data edges, nearby examples with different ground-truth labels, and model edges, similar examples previously misclassified by the deployed model. EdgeLM requires neither model retraining nor task-specific engineering. Across five data wrangling tasks, fifteen datasets, and five open-weight and proprietary LLMs, EdgeLM consistently achieves the best or near-best performance in every setting, while ablations show that the two forms of edge evidence provide complementary benefits. Our code and datasets are publicly available at https://github.com/soroushomidvar/EdgeLM.
