 |
ML Newsnew | past | best | rss | submit | about |
| Hallucination Stations: Limitations of Transformer-Based Language Models (2025) (arxiv.org) | |
| 1 point by cainxinth 253 days ago | hide | past | pdf | discuss on HN | |
|
In plain words: Using ideas from computational complexity, the paper shows that once tasks pass a certain level of complexity, transformer-based language models cannot complete them or check whether their answers are correct. This sets a hard ceiling on what such models and agents can reliably do, including verifying their own work. Abstract · Hallucination Stations: On Some Basic Limitations of Transformer-Based Language ModelsIn this paper we explore hallucinations and related capability limitations in LLMs and LLM-based agents from the perspective of computational complexity. We show that beyond a certain complexity, LLMs are incapable of carrying out computational and agentic tasks or verifying their accuracy. | |
| add comment on HN |