Every story in this issue with a short summary, so you can decide what to read before leaving the site. Summaries are written by AI (Claude) only from each publisher's own description and checked by Jev for anything not in that description. They are not full translations. Follow "Read the original" for the full story.
A new study reports a bumble bee mating-optimized convolutional neural network that classifies seven student emotional states from facial expressions. It achieved 98.25 percent test accuracy.
Read the original →A new contrastive learning framework adapts a speech-trained HuBERT model to unlabeled violin recordings. It achieved the best reported correlations with perceived emotion on the VioMusic benchmark.
Read the original →NTT Docomo has developed an evaluation framework, a reference architecture and four supporting technologies to help conversational AI respond more appropriately to users' emotions and circumstances while maintaining safety. The AI EQ+1 Framework assesses responses across four areas: basic performance, safety and integrity, interaction quality, and user outcomes.
Read the original →NTT announced on October 27, 2026 a high-EQ dialogue intelligence technology that curbs sycophancy, where AI over-agrees with users, and considers long-term real-life benefit rather than short-term satisfaction. On a public EQ test, scores rose about 23% compared with before the technology was applied. NTT plans trials toward implementation in smartphone chat services.
Read the original →The connected coaching experience pairs a dedicated ICF-certified coach with Cai, an AI development partner. Cai remembers conversation content, priorities and organizational context, and is intended to help managers improve their performance.
Read the original →