PaperMind is a new benchmark that evaluates integrated multimodal reasoning and critique over scientific papers through four complementary task families across seven domains.
arXiv preprint arXiv:2503.16856 , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.IR 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
PaperMind: Benchmarking Agentic Reasoning and Critique over Scientific Papers in Multimodal LLMs
PaperMind is a new benchmark that evaluates integrated multimodal reasoning and critique over scientific papers through four complementary task families across seven domains.