Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
4
2
Contextual coreference models aren't always better: simple lexical matching can be more robust to mention substitution noise, while contextual models shine with noisy boundaries.
Vocabulary-constrained LLMs can rival domain-adapted models in astrophysics classification, hinting at parameter-efficient strategies for scientific NLP.