Yiderigun Borjigin

AnchorBench: A Multi-Pathway Benchmark for the Anchoring Effect in LLMs

The anchoring effect is a cognitive bias in which an initial reference value shifts a later judgment toward itself. This effect is well established in human judgment and decision-making, and recent work suggests that large language models (LLMs) exhibit similar behavior. However, existing work on anchoring in LLMs typi...