Scaling of RLHF-Blender with Model Size in HumanEval-plus Pass@k Performance

Abstract

Abstract: We apply preference modeling and reinforcement learning from human feedback (RLHF) to netune language models to act as helpful and harmless assistants. We nd this alignment training improves performance on almost all NLP evaluations, and is fully compatible with training for specialized skills such as python coding and summarization. We explore an iterated online mode of training, where preference models and RL policies are updated on a weekly cadence with fresh human feedback data, efciently improving our datasets and models. Finally, we investigate the robustness of RLHF training, and ide

Research Question

How does the RLHF-Blender approach scale with increasing model size in terms of pass@k performance on HumanEval-plus compared to independent sampling in CodeT5+?

Verification Level

Descriptive public verification status only; aggregate claim counts are public, but individual claim records are not exposed here.

Truth-Engine Gate Verdict

This record has not completed Gate 2 of the verification pipeline (a type-checked Lean4 proof for mathematical claims, or a sealed-sandbox reproduction for empirical claims). It is a literature synthesis only. VERIFIED requires an attached reproducible artifact (Lean4 proof source, or repro script and results) before this status can be set; it is not derived from review score or claim count.

Quality Tier

Descriptive public triage only; this tier does not alter current publication or DOI behavior.

Quality Dimensions

Automated triage signals derived from public fields; not human peer review or independent validation.

Correction Record

Public corrections are additive records. Current status does not claim the synthesis is error-free.

Provenance