HiddenBench
PulseAugur coverage of HiddenBench — every cluster mentioning HiddenBench across labs, papers, and developer communities, ranked by signal.
-
New model explains how LLM agents can reach incorrect consensus
A new model explains how multi-agent systems of LLMs can sometimes reach incorrect consensus due to agents withholding dissent. The model identifies a critical withholding rate, below which discussion improves accuracy.…
-
Consilience framework enhances multi-agent LLM communication and decision-making
Researchers have introduced Consilience, a new framework designed to improve communication and decision-making in multi-agent large language model (LLM) systems. This framework addresses challenges in hidden-profile sce…
-
New LLM tool detects hidden disagreements in collaborative AI dialogues
Researchers have developed a method to detect "illusion of alignment" (IoA) in collaborative dialogues, where participants appear to agree but hold differing underlying goals or assumptions. They created IoA-Suite, a da…