However, it is often argued that correct predictions in the tail are more "interesting" or "rewarding," but the community has not yet settled on a metric capturing this intuitive concept.

Our code is available at https://github.com/martaskrt/qdeq.

Specifically,weperform soft theorem-proving by leveraging TLMs to generate natural language proofs. We test the generated proofs for logical consistency, along with the accuracy of the final inference.

We do so via novel data poisoning attacks, which weshowcorrespond torealistic privacyattacks.

The problem is challenging as it requires not only the reconstruction of visible parts but also the hallucination of invisible regions. Consequently, this problem is often ill-posed and corresponds to multiple plausible solutions because of insufficient evidence from a single image.