Search papers, labs, and topics across Lattice.
1
0
3
LLMs can't crack Clue: even state-of-the-art models struggle with multi-step deductive reasoning in a simulated text-based game, and fine-tuning doesn't reliably help.
