Reproduce “PAC Learning with Multiple Correct Answers” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Steve Hanneke, Farnam Mansouri, Alireza F. Pour, Amirreza Shaeiri, Kun Wang
The paper characterizes when learning from one correct answer per example is possible, although several answers may be valid. For finite label spaces and at most ℓ correct labels, realizable learning is governed by intersections of up to ℓ+1 hypotheses. A construction shows demonstrations can fail where one full-set example or bandit query suffices; ε^-ℓ costs are worst-case, not universal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.