SocialHarmBench Uncovers AI Vulnerabilities in Sociopolitical Contexts
SocialHarmBench, a new benchmark with 585 prompts across 34 countries, shows that the open‑weight Mistral‑7B model complies with harmful requests at a 97‑98% success rate. getnews.me/socialharmbench-uncovers... #socialharmbench #mistral7b
Teaching Language Models to Capture Human Word Choice Variability
Fine‑tuning GPT‑2 and Mistral‑7B‑IT on the Provo Corpus with a multi‑label loss reduced KL‑divergence from human word‑choice distributions, boosting results in high‑variability contexts. Read more: getnews.me/teaching-language-models... #gpt2 #mistral7b
Open-Source LLMs Benchmarked for Multi-Intent Dialogue Classification
Mistral‑7B‑v0.1 posted a weighted F1 of 0.50 on the MultiWOZ 2.1 multi‑intent benchmark, but a fine‑tuned BERT classifier still outperformed all few‑shot models. Read more: getnews.me/open-source-llms-benchma... #mistral7b #bert #dialogue
Run open-weight #LLMs like #Mistral7B on your own terms. 🧠
Our latest guide shows how to use #Ollama + #UpCloud GPUs for private, high-perf AI inference.
Tutorial → upcloud.com/resources/tu...
#OpenSourceAI #CloudComputing
@MistralAI, étoile montante dans le domaine de l’IA en France, veut faire aussi bien qu’@OpenAI. En septembre dernier, elle dévoilait #Mistral7B, son premier grand modèle open source pour rivaliser avec les acteurs américains. Mais le challenge est de taille.
#BDM #AyTèk
