R0040/2026-04-01/Q002/SRC01
Shapira, Benade, Procaccia -- How RLHF Amplifies Sycophancy (2026)
Source
| Field |
Value |
| Title |
How RLHF Amplifies Sycophancy |
| Publisher |
arXiv |
| Author(s) |
Itai Shapira, Gerdus Benade, Ariel D. Procaccia |
| Date |
2026-02-01 |
| URL |
https://arxiv.org/abs/2602.01002 |
| Type |
Research paper |
Summary
| Dimension |
Rating |
| Reliability |
High |
| Relevance |
High |
| Bias: Missing data |
Low risk |
| Bias: Measurement |
Low risk |
| Bias: Selective reporting |
Low risk |
| Bias: Randomization |
N/A -- not an RCT |
| Bias: Protocol deviation |
N/A -- not an RCT |
| Bias: COI/Funding |
Low risk |
Rationale
| Dimension |
Rationale |
| Reliability |
Formal mathematical analysis from established researchers (Procaccia is a well-known computational social choice theorist). Provides proofs, not just experiments. |
| Relevance |
Directly proves the mechanism by which RLHF amplifies sycophancy. Most relevant source for Q002. |
| Bias flags |
No identified conflicts. The research is academic without commercial stake. |
| Evidence ID |
Summary |
| SRC01-E01 |
Formal proof: RLHF amplifies sycophancy via covariance between agreement and reward; proposed reward correction |