Back to Research papers
Research paper index

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

Yiderigun Borjigin, Alexander Hermann, Christian Cyron, Roland Aydin

arXiv:2608.14320Published August 14, 20260 citations
  • cs.AI
  • cs.CL

Abstract

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 typically evaluates only a narrow set of anchor pathways and rarely distinguishes irrelevant from plausible anchors. We introduce AnchorBench, a benchmark for the anchoring effect in LLMs that evaluates multiple anchor pathways under an explicit anchor relevance axis. Across fourteen models, including ten open-weight models and four frontier API models, and a large set of controlled prompts, we find that (1) anchoring is strongly pathway-dependent, (2) plausible anchors usually induce larger shifts than irrelevant ones when introduced through stronger pathways, (3) anchor influence generally weakens as the anchor moves farther from the evidence-supported answer, most clearly on External and RAG, and (4) high task accuracy on the anchor-free control condition (Acc$_{10}$: answers within 10 points of gold) does not guarantee robustness: even frontier API models above 95% control accuracy remain susceptible to plausible anchors.

Read the original paper

This page indexes public paper metadata. The manuscript remains with its original publisher and authors.