Search papers, labs, and topics across Lattice.
5
5
6
13
Bias detection in neural networks can be revolutionized by analyzing latent spaces and activations, revealing how biases are embedded in model architecture itself.
Visual understanding, not knowledge, is the critical barrier in Document Visual Question Answering, with smaller models showing surprising adaptability through targeted finetuning.
Current AI risk assessment practices are inadequate, revealing critical gaps that could jeopardize the safety of intelligent systems in a rapidly evolving regulatory environment.
A lightweight model fine-tuned on AIriskEval-edu-db2 can rival leading models in pedagogical risk detection, all while maintaining privacy in educational settings.
CrimeNERdb offers a crucial resource for improving crime-related information extraction, filling a significant gap in NER datasets.
