E01¶


Research	R0055 — RLHF Yes-Men Claims
Run	2026-04-01
Claim	C009
Source	SRC01
Evidence	SRC01-E01
Type	Analytical

RLVR primarily works in math/code but active research extends it to other domains; 'only works' is overstated

URL: https://arxiv.org/pdf/2503.23829

Extract¶

RLVR has mainly demonstrated success on tasks with precisely structured solutions such as mathematical reasoning or code generation. However, 'only works' overstates the limitation: research is expanding RLVR to knowledge-intensive domains. Only 60.3% of math problems have verifiable single-term answers. RLVR fails for creative writing, brand voice, or nuanced argumentation.

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Supports	Moderate
H2	Supports	Strong
H3	Contradicts	Strong

Context¶

Evidence directly relevant to testing the claim's factual assertions.