C007 — Claim Definition¶


Research	R0055 — RLHF Yes-Men Claims
Run	2026-04-01
Claim	C007

Claim as Received¶

Six major alternatives to RLHF have emerged since 2022 (DPO, Constitutional AI, GRPO, KTO, ORPO, RLVR)

Claim as Clarified¶

Six major alternatives to RLHF have emerged since 2022 (DPO, Constitutional AI, GRPO, KTO, ORPO, RLVR)

BLUF¶

Substantially correct. All six named methods exist and have emerged as alternatives or complements to standard RLHF. DPO (2023), Constitutional AI (2022), GRPO (2025), KTO (2024), ORPO (2024), and RLVR (2024-2025) are all documented. Whether they are all 'major' is debatable — some like DPO and GRPO are widely adopted while others like KTO and ORPO have narrower use.

Scope¶

Domain: AI alignment, sycophancy, enterprise AI
Timeframe: 2022-2026
Testability: Verifiable against published research and documentation

Assessment Summary¶

Probability: Very likely (80-95%)

Confidence: High

Hypothesis outcome: H2 prevails — see assessment for details.

[Full assessment in assessment.md.]

Status¶

Field	Value
Date created	2026-04-01
Date completed	2026-04-01
Researcher profile	Phillip Moore
Prompt version	Unified Research Methodology v1
Revisit by	2026-10-01
Revisit trigger	New methods displacing any of these six; changes in adoption patterns