AI is stepping into the arena to tackle online hate speech and misinformation. A new approach combines expert guidelines with AI's ability to generate counterspeech.
DeferMem introduces a novel approach to handling long-term memory in AI models, focusing on query-conditioned evidence distillation. It promises enhanced accuracy and efficiency without the cost of commercial APIs.
High-Entropy Sum (HES) emerges as a training-free metric boosting language models' reasoning capabilities. By focusing on high-entropy tokens, it cuts computational costs while enhancing model performance.
