Skip to content

R0056/2026-04-01/C009/SRC01/E01

Research R0056 — RLHF Yes-Men Claims v2
Run 2026-04-01
Claim C009
Source SRC01
Evidence SRC01-E01
Type Reported

Primary evidence for C009

URL: See source scorecard

Extract

Anthropic describes sycophancy as a simple form of specification gaming that can escalate to more sophisticated behaviors like reward tampering.

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Supports See assessment
H2 Supports See assessment
H3 Contradicts See assessment

Context

See assessment.md for full context.