Qwen1.5
PulseAugur coverage of Qwen1.5 — every cluster mentioning Qwen1.5 across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New probe DANGERLENS reveals hidden dangerous concepts in AI models
Researchers have developed DANGERLENS, a new probe designed to uncover hidden dangerous concepts within language models that may persist despite safety alignment training. By mapping internal model states to vocabulary …
-
Chinese open-weight LLM release timeline compiled from vendor data · 1 source tracked
A comprehensive timeline details the release dates of open-weight large language models from Chinese organizations, meticulously compiled from vendor pages, research papers, and Hugging Face's API. The list excludes API…
-
MoE language models show limited stereotype control via routing
A new research paper explores methods for controlling stereotypes in Mixture-of-Experts (MoE) language models by adjusting routing mechanisms. The study introduces the Fairness-Aware Routing Equilibrium (FARE) framework…
-
Alibaba to charge enterprises for its next open-source AI model
Alibaba is reportedly planning to monetize its upcoming open-source AI model, Qwen1.5. While the model will be available for free for smaller users, larger enterprises are expected to incur charges. This move signals a …
-
Open-source LLM efficient frontier charted by parameter efficiency
A Reddit user has compiled a chart illustrating the efficient frontier of open-source large language models, defining efficiency as the model's score relative to its active parameters. The chart focuses on models that r…