Tag
This paper evaluates a multimodal framework for speaker identification in K-12 classrooms by combining acoustic embeddings (ECAPA-TDNN) with LLM-derived semantic context from transcripts, improving accuracy from 39% to 50.3% overall and from 64.9% to 76.9% for longer utterances.
Introduces LiveK12Bench, a dynamic multimodal benchmark for evaluating large models on real high school exams across multiple disciplines, revealing substantial performance gaps under realistic constraints.
OpenAI is launching ChatGPT for Teachers, a free workspace available through June 2027 that includes GPT-4.1 Auto, education-grade privacy/FERPA compliance, collaboration tools, and integrations with Google Drive, Microsoft 365, and Canva, aimed at K-12 educators and school administrators in the U.S.
OpenAI and the American Federation of Teachers launch the National Academy for AI Instruction, a five-year initiative to train 400,000 K-12 educators (about 1 in 10 US teachers) in AI usage and integration in classrooms. OpenAI is contributing $10 million in funding and resources, with support from Microsoft, Anthropic, and the United Federation of Teachers.