Year: 2026
Type: preprint
Abstract: Language model post-training is often bottlenecked by the need for human-collected preference data, which is expensive and difficult to scale. Reinforcement learning from AI feedback (RLAIF) style approaches that leverage pseudo labels offer an abundant alternative but introduce systematic biases that degrade downstream alignment. Recent general-purpose semi-supervised methods correct for teacher bias using a small set of human-labeled examples, but suffer from high variance especially when human annotations are scarce. To this end, we propose ABC-Align, leveraging abundant pseudo label signal to minimize variance and applying a lightweight, adaptive correction grounded in the human-labeled subset. The correction strength is tuned automatically during training using plug-in estimates of the relevant bias--variance quantities. On LLM alignment with RLHF, DPO, and GRPO where human feedback is scarce, we empirically demonstrate that ABC-Align achieves superior performance over prior semi-supervised baselines in a series of experiments on an increasing scale. Our code is available at https://github.com/SewoongLab/abc-align .
Source: arXiv (Cornell University)
Institution: University of Washington
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: National Science Foundation
Awards: AI Institute for Future Edge Networks and Distributed Intelligence (AI-EDGE), Collaborative Research: MFAI: Mathematical Frontiers of Generative AI, Collaborative Research: AF: Medium: Machine Learning Markets: Dynamics, Competition, and Interventions, AI Institute for Agent-based Cyber Threat Intelligence and Operation, Institute for Foundations of Machine Learning
OpenAlex ID: W7214798453
Cites:
Cited by:
Related to: