Search papers, labs, and topics across Lattice.
Affiliation:
3
0
6
Current evaluations of coding agents may mismeasure performance by conflating action, task, and step levels, revealing a critical flaw in how we assess agent execution.
DeepDiscovery boosts task-relevant file recovery by up to 9.2 percentage points, transforming how coding agents navigate complex industrial codebases.
Combat noisy click data in recommender systems by simply reweighting training samples based on the semantic similarity between user interest profiles and item descriptions, yielding surprisingly robust performance gains.
