BACH-V: Bridging Abstract and Concrete Human-Values in Large Language Models
Junyu Zhang, Yipeng Kang, Jiong Guo, Jiayu Zhan, Junqi Wang · 2026
Read BACH-V: Bridging Abstract and Concrete Human-Values in Large Language Models by Junyu Zhang, Yipeng Kang, Jiong Guo, Jiayu Zhan, Junqi Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-14007.txt.
Description
Junyu Zhang and co-authors test if language models truly grasp abstract human values in 2026. A model might handle a value word as a mere statistical pattern. Or it might link the word to real events and choices. Their BACH-V framework splits conceptual understanding into three capacities.
One capacity interprets abstract concepts, and a second grounds them in concrete events. The third applies abstract principles to regulate concrete decisions. The names are Abstract-Abstract, Abstract-Concrete, and Concrete-Concrete. Testing each layer shows where a model's grasp of values breaks down. The split into three tests makes flaws in a model easy to spot. Each test can be run on its own with plain tasks.
Catalog record
| CREATOR | Junyu Zhang, Yipeng Kang, Jiong Guo, Jiayu Zhan, Junqi Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 85 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-14007