Search papers, labs, and topics across Lattice.
2
0
4
Prompt robustness varies dramatically between objective and subjective questions, revealing critical vulnerabilities in how we interpret LLM responses.
By explicitly learning to filter soft noise with a physics-constrained gate, PhyGHT achieves state-of-the-art signal purification in simulated HL-LHC data, outperforming existing methods from ATLAS and CMS.
