Centaur, a foundation model trained on 160 general psychological experiments, is evaluated and it is found that Centaur more closely aligns with human response patterns than its base model, is significantly less reliant on information from prior trials and responses, and benefits more from task-related information.
The findings show that semantic conflicts affect program comprehension and downstream tasks, with relevant information concentrated in a small number of causally active residual-stream states, and demonstrate a framework for mechanistically analyzing how LLMs integrate code-related information under controlled semantic...
Youssef Abdelsalam, Norman Peitek, Anna-Maria Maurer et al.· arXiv.org· 2 citations
This work introduces a facet-based perspective that distinguishes between analytical skills, abstraction, and critical evaluation and proposes a method to guide researchers from research goals to concrete tasks and study design, including considerations for question formulation and response formats.
Luisa Bartl, Marvin Wyrich, Sven Apel et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
