Search papers, labs, and topics across Lattice.
2
0
4
0
Unified multimodal models can ace visual understanding and generation tasks, yet still fail to maintain basic semantic consistency between them.
Injecting emotion-aware features derived from a DistilBERT-expanded lexicon into a neural network boosts argumentative stance classification, outperforming even LLM-based methods on controversial topics.
