Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Kellogg School of Management, Center for Science of Science and Innovation, Northwestern Innovation Institute, Ryan Institute on Complexity, and McCormick School of Engineering, Northwestern University, dashun.wang@northwestern.edu
1
0
3
0
Don't trust the benchmark: Seemingly equivalent LLMs can disagree wildly on individual examples, leading to irreproducible scientific results when used for annotation.