Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2503.11656 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910876707061760 |
|---|---|
| author | Liu, Joshua Jain, Aarav Takuri, Soham Vege, Srihan Akalin, Aslihan Zhu, Kevin O'Brien, Sean Sharma, Vasu |
| author_facet | Liu, Joshua Jain, Aarav Takuri, Soham Vege, Srihan Akalin, Aslihan Zhu, Kevin O'Brien, Sean Sharma, Vasu |
| contents | Rapid improvements in large language models have unveiled a critical challenge in human-AI interaction: sycophancy. In this context, sycophancy refers to the tendency of models to excessively agree with or flatter users, often at the expense of factual accuracy. While previous studies have primarily analyzed this behavior in single-turn interactions, its persistence and evolution in multi-step conversations remain largely unexplored. We introduce TRUTH DECAY, a benchmark specifically designed to evaluate sycophancy in extended dialogues, where language models must navigate iterative user feedback, challenges, and persuasion. We prompt models to elicit four types of sycophantic biases. We then propose and test sycophancy reduction strategies, evaluating their effectiveness beyond single-step interactions. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_11656 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TRUTH DECAY: Quantifying Multi-Turn Sycophancy in Language Models Liu, Joshua Jain, Aarav Takuri, Soham Vege, Srihan Akalin, Aslihan Zhu, Kevin O'Brien, Sean Sharma, Vasu Computation and Language Rapid improvements in large language models have unveiled a critical challenge in human-AI interaction: sycophancy. In this context, sycophancy refers to the tendency of models to excessively agree with or flatter users, often at the expense of factual accuracy. While previous studies have primarily analyzed this behavior in single-turn interactions, its persistence and evolution in multi-step conversations remain largely unexplored. We introduce TRUTH DECAY, a benchmark specifically designed to evaluate sycophancy in extended dialogues, where language models must navigate iterative user feedback, challenges, and persuasion. We prompt models to elicit four types of sycophantic biases. We then propose and test sycophancy reduction strategies, evaluating their effectiveness beyond single-step interactions. |
| title | TRUTH DECAY: Quantifying Multi-Turn Sycophancy in Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2503.11656 |