Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
Post-training LLMs for agentic tasks can drastically reduce solution coverage, a "Sharpening Tax" that a novel Bayesian sampler can mitigate.
