Search papers, labs, and topics across Lattice.
This paper introduces D2VBench, a novel benchmark designed to evaluate large language models (LLMs) on their alignment with human values through 10,000 instances of real-world dilemma scenarios. The benchmark addresses the limitations of existing evaluation methods by incorporating a hybrid evaluation paradigm that combines multiple-choice and open-ended questions, allowing for a more nuanced assessment of value conflicts. Experimental results show that D2VBench reliably captures LLMs' alignment across various value dimensions, making it a significant advancement in the study of value alignment in AI systems.
LLMs are assessed on their alignment with human values through a groundbreaking benchmark that captures real-world dilemmas, revealing critical insights into their decision-making processes.
With the wide application of large language models (LLMs) in real-world scenarios, the value implication of their outputs is crucial. However, existing evaluation benchmarks suffer from insufficient coverage of value dilemmas in daily scenarios involving multiple value conflicts and simplistic evaluation formalisms that fail to assess LLMs'value alignment. To address these issues, we propose D2VBench, a value alignment benchmark comprising 10,000 instances of real daily dilemma scenarios constructed through a multi-stage collaboration between LLMs and humans, grounded in 158 manually annotated fine-grained value concepts. For evaluation on the benchmark, we present a hybrid evaluation paradigm that integrates multiple-choice questions with open-ended questions. We conduct comprehensive evaluations on eight mainstream LLMs. Experimental results demonstrate that D2VBench exhibits high reliability and robustness, effectively reflecting the LLMs'alignment across different value categories and dimensions, and providing a more realistic and fine-grained tool for research on value alignment. The dataset is available at https://github.com/tjunlp-lab/D2VBench.
