Atomic Calibration of LLMs in Long-Form Generations
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908666123255808 |
|---|---|
| author | Zhang, Caiqi Yang, Ruihan Zhang, Zhisong Huang, Xinting Yang, Sen Yu, Dong Collier, Nigel |
| author_facet | Zhang, Caiqi Yang, Ruihan Zhang, Zhisong Huang, Xinting Yang, Sen Yu, Dong Collier, Nigel |
| contents | Large language models (LLMs) often suffer from hallucinations, posing significant challenges for real-world applications. Confidence calibration, as an effective indicator of hallucination, is thus essential to enhance the trustworthiness of LLMs. Prior work mainly focuses on short-form tasks using a single response-level score (macro calibration), which is insufficient for long-form outputs that may contain both accurate and inaccurate claims. In this work, we systematically study atomic calibration, which evaluates factuality calibration at a fine-grained level by decomposing long responses into atomic claims. We further categorize existing confidence elicitation methods into discriminative and generative types, and propose two new confidence fusion strategies to improve calibration. Our experiments demonstrate that LLMs exhibit poorer calibration at the atomic level during long-form generation. More importantly, atomic calibration uncovers insightful patterns regarding the alignment of confidence methods and the changes of confidence throughout generation. This sheds light on future research directions for confidence estimation in long-form generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_13246 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Atomic Calibration of LLMs in Long-Form Generations Zhang, Caiqi Yang, Ruihan Zhang, Zhisong Huang, Xinting Yang, Sen Yu, Dong Collier, Nigel Computation and Language Artificial Intelligence Large language models (LLMs) often suffer from hallucinations, posing significant challenges for real-world applications. Confidence calibration, as an effective indicator of hallucination, is thus essential to enhance the trustworthiness of LLMs. Prior work mainly focuses on short-form tasks using a single response-level score (macro calibration), which is insufficient for long-form outputs that may contain both accurate and inaccurate claims. In this work, we systematically study atomic calibration, which evaluates factuality calibration at a fine-grained level by decomposing long responses into atomic claims. We further categorize existing confidence elicitation methods into discriminative and generative types, and propose two new confidence fusion strategies to improve calibration. Our experiments demonstrate that LLMs exhibit poorer calibration at the atomic level during long-form generation. More importantly, atomic calibration uncovers insightful patterns regarding the alignment of confidence methods and the changes of confidence throughout generation. This sheds light on future research directions for confidence estimation in long-form generation. |
| title | Atomic Calibration of LLMs in Long-Form Generations |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2410.13246 |