Evaluating the Paperclip Maximizer: Are RL-Based Language Models More Likely to Pursue Instrumental Goals?

Y He, Y Li, J Wu, Y Sui, Y Chen, B Hooi - arxiv preprint arxiv:2502.12206, 2025 - arxiv.org
As large language models (LLMs) continue to evolve, ensuring their alignment with human
goals and values remains a pressing challenge. A key concern is\textit {instrumental …

Structured Knowledge meets GenAI: A Framework for Logic-Driven Language Models

FH Eldessouky, N Ehab, C Schindler… - Proceedings of the …, 2025 - aclanthology.org
Abstract Large Language Models (LLMs) excel at generating fluent text but struggle with
context sensitivity, logical reasoning, and personalization without extensive fine-tuning. This …