Robustness of DPO and RLHF Alignment Under Varying Dataset Sizes in Multimodal Reasoning

Abstract

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the impact of dataset size on the robustness of DPO versus RLHF alignment methods when evaluated on multimodal reasoning benchmarks with corrupted image-text pairs. This paper studies the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We first identify the primary challenges of existing popular methods like offline PPO and offline DPO as lacking in strategical exploration of the environment. 9 claims were extracted from source literature; 0 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 3.3/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research Question

What is the impact of dataset size on the robustness of DPO versus RLHF alignment methods when evaluated on multimodal reasoning benchmarks with corrupted image-text pairs?

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