Enhancing Hallucination Detection via Future Context
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910106250117120 |
|---|---|
| author | Lee, Joosung Park, Cheonbok Jo, Hwiyeol Kim, Jeonghoon Park, Joonsuk Yoo, Kang Min |
| author_facet | Lee, Joosung Park, Cheonbok Jo, Hwiyeol Kim, Jeonghoon Park, Joonsuk Yoo, Kang Min |
| contents | Large Language Models (LLMs) are widely used to generate plausible text on online platforms, without revealing the generation process. As users increasingly encounter such black-box outputs, detecting hallucinations has become a critical challenge. To address this challenge, we focus on developing a hallucination detection framework for black-box generators. Motivated by the observation that hallucinations, once introduced, tend to persist, we sample future contexts. The sampled future contexts provide valuable clues for hallucination detection and can be effectively integrated with various sampling-based methods. We extensively demonstrate performance improvements across multiple methods using our proposed sampling approach. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_20546 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Enhancing Hallucination Detection via Future Context Lee, Joosung Park, Cheonbok Jo, Hwiyeol Kim, Jeonghoon Park, Joonsuk Yoo, Kang Min Computation and Language Artificial Intelligence Large Language Models (LLMs) are widely used to generate plausible text on online platforms, without revealing the generation process. As users increasingly encounter such black-box outputs, detecting hallucinations has become a critical challenge. To address this challenge, we focus on developing a hallucination detection framework for black-box generators. Motivated by the observation that hallucinations, once introduced, tend to persist, we sample future contexts. The sampled future contexts provide valuable clues for hallucination detection and can be effectively integrated with various sampling-based methods. We extensively demonstrate performance improvements across multiple methods using our proposed sampling approach. |
| title | Enhancing Hallucination Detection via Future Context |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2507.20546 |