Red Teaming for Large Language Models At Scale: Tackling Hallucinations on Mathematics Tasks
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917557371404288 |
|---|---|
| author | Buszydlik, Aleksander Dobiczek, Karol Okoń, Michał Teodor Skublicki, Konrad Lippmann, Philip Yang, Jie |
| author_facet | Buszydlik, Aleksander Dobiczek, Karol Okoń, Michał Teodor Skublicki, Konrad Lippmann, Philip Yang, Jie |
| contents | We consider the problem of red teaming LLMs on elementary calculations and algebraic tasks to evaluate how various prompting techniques affect the quality of outputs. We present a framework to procedurally generate numerical questions and puzzles, and compare the results with and without the application of several red teaming techniques. Our findings suggest that even though structured reasoning and providing worked-out examples slow down the deterioration of the quality of answers, the gpt-3.5-turbo and gpt-4 models are not well suited for elementary calculations and reasoning tasks, also when being red teamed. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_00290 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Red Teaming for Large Language Models At Scale: Tackling Hallucinations on Mathematics Tasks Buszydlik, Aleksander Dobiczek, Karol Okoń, Michał Teodor Skublicki, Konrad Lippmann, Philip Yang, Jie Computation and Language Artificial Intelligence I.2.7 We consider the problem of red teaming LLMs on elementary calculations and algebraic tasks to evaluate how various prompting techniques affect the quality of outputs. We present a framework to procedurally generate numerical questions and puzzles, and compare the results with and without the application of several red teaming techniques. Our findings suggest that even though structured reasoning and providing worked-out examples slow down the deterioration of the quality of answers, the gpt-3.5-turbo and gpt-4 models are not well suited for elementary calculations and reasoning tasks, also when being red teamed. |
| title | Red Teaming for Large Language Models At Scale: Tackling Hallucinations on Mathematics Tasks |
| topic | Computation and Language Artificial Intelligence I.2.7 |
| url | https://arxiv.org/abs/2401.00290 |