masked language models
PulseAugur coverage of masked language models — every cluster mentioning masked language models across labs, papers, and developer communities, ranked by signal.
-
New pipeline debiases phrase-level stereotypes in language models
Researchers have developed a "General Phrase Debiaser," a novel pipeline designed to mitigate phrase-level biases in masked language models. This method identifies stereotypical phrases from sources like Wikipedia and t…
-
Language model semantic spaces show varied representation of word relations
Researchers have investigated the geometric representation of semantic relations within the vector spaces of language models. Their study explored whether words related to a target word occupy similar regions and if the…
-
AI models advance ad headline generation with improved CTR and quality · 2 sources tracked
Two new research papers propose advanced methods for generating advertising headlines. One paper introduces COBART, a controlled, optimized, bidirectional, and auto-regressive Transformer model that uses prefix control …
-
New framework unifies in-context learning analysis for causal and masked models
Researchers have developed a unified statistical learning framework to analyze in-context learning (ICL) capabilities in both causal and masked language models. This framework places autoregressive and masked pretrainin…