Search papers, labs, and topics across Lattice.
Anderson Avila, Institut national de la recherche scientifique, Québec, Montréal, Canada
1
0
2
1
Ideological biases in RAG can significantly shape LLM outputs, with moderate sampling temperatures amplifying discourse alignment.
