Skip to content

R0055/2026-04-01/C006/H2

Research R0055 — RLHF Yes-Men Claims
Run 2026-04-01
Claim C006
Hypothesis H2

Statement

Claim is partially correct or correct with caveats

Status

Current: Inconclusive

Supporting Evidence

Evidence Summary
SRC01-E01 Synthetic data achieved 4.7-10% sycophancy reduction, far less than 84-85% from curated pairs

Contradicting Evidence

Evidence Summary
— No contradicting evidence identified

Reasoning

This hypothesis remains inconclusive based on available evidence.

Relationship to Other Hypotheses

H2 is secondary to the supported hypothesis.