Search papers, labs, and topics across Lattice.
2
0
5
Despite impressive headline accuracy, today's AI chatbots exhibit alarming regional biases, near-total dependence on retrieval quality, and surprising vulnerability to subtle falsehoods in user queries when used as news intermediaries.
Uncover the surprisingly small fraction of model parameters (as low as 2.4% of adapter features) responsible for specific reasoning behaviors like hesitation token generation, offering a path to targeted model editing.
