Oracle-RLAIF CIDEr Gains Over SFT vs Reinforcement Learning Baselines on MSVD

Abstract

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the CIDEr score improvement of Oracle-RLAIF over SFT compare to other reinforcement learning methods (e.g., PPO, DQN) on the MSVD benchmark across different model sizes. In post-training for reasoning Large Language Models (LLMs), the current state of practice trains LLMs in two independent stages: Supervised Fine-Tuning (SFT) and Reinforcement Learning with Verifiable Rewards (RLVR, shortened as ``RL'' below). In this work, we challenge whether. 6 claims were extracted from source literature; 0 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 4.5/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research Question

How does the CIDEr score improvement of Oracle-RLAIF over SFT compare to other reinforcement learning methods (e.g., PPO, DQN) on the MSVD benchmark across different model sizes?

Verification Level

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

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