Search papers, labs, and topics across Lattice.
Carnegie Mellon University
1
0
3
12
Even GPT-5 only achieves 63% accuracy on time series anomaly questions from real software incidents, but a model-expert combination reaches 87%, highlighting the potential for hybrid intelligence in incident response.
