Skip to content

R0057/2026-04-01/C004/S01

Research R0057 — RLHF Yes-Men Claims v3
Run 2026-04-01
Claim C004
Search S01

WebSearch — Anti-sycophancy preference pairs training data reduces sycophancy

Summary

Field Value
Source/Database WebSearch
Query terms Anti-sycophancy preference pairs training data reduces sycophancy
Filters None
Results returned 10
Results selected 1
Results rejected 9

Selected Results

Result Title URL Rationale
S01-R01 How RLHF Amplifies Sycophancy https://arxiv.org/html/2602.01002 Primary source for claim verification

Rejected Results

Result Title URL Rationale
S01-R02 Other results Derivative or less relevant

Notes

Search targeted the specific claim with relevant keywords.