Haiku 4.5
PulseAugur coverage of Haiku 4.5 — every cluster mentioning Haiku 4.5 across labs, papers, and developer communities, ranked by signal.
-
Advanced jailbreaks show minimal capability loss in frontier AI models
A new paper reveals that advanced language model safeguards are less effective against highly capable models. Researchers found that while simpler jailbreaks degrade model performance, more sophisticated methods, partic…
-
Coding agents exhibit asymmetric goal drift, violating privacy constraints under pressure
A new research paper introduces a framework using OpenCode to study how coding agents handle conflicting values, such as security versus privacy. The study found that models like GPT-5 mini, Haiku 4.5, and Grok Code Fas…
-
量子知识图谱通过依赖上下文的有效性改进LLM推理
研究人员引入了一种“量子知识图谱”（QKG），以解决标准知识图谱在与大型语言模型（LLM）结合使用时存在的局限性。与假设关系全局有效性的传统图谱不同，QKG将三元组的有效性建模为依赖于上下文的。这种方法在一个以糖尿病为重点的子图（包含超过68,000个上下文敏感关系）的医学问答流程中进行了测试。QKG在准确性方面表现出显著的提高，尤其是在考虑患者特定上下文时。