Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
3
0
6
8
Pruning LLMs can amplify biases, but Debias-SparseGPT shows that you can reduce these biases while maintaining performance, even under aggressive sparsity constraints.
Fairness degradation in language models can silently escalate before performance metrics signal any issues, revealing a hidden danger in synthetic data training.
LLMs can achieve fairer outcomes through multi-agent negotiation than through individual ethical alignment, suggesting a path to fairness via interaction rather than monolithic design.