ReNCE: Learning to Reason by Noise Contrastive Estimation
This work addresses the challenge of improving reasoning in LLMs for tasks like math problem-solving, but it is incremental as it builds on existing contrastive learning and reasoning methods.
The paper tackles the problem of endowing pretrained LLMs with reasoning capabilities by proposing ReNCE, an explicit contrastive learning approach that bifurcates outcomes into positive and negative sets to maximize likelihood, instead of using the standard GRPO method with empirical refinements. It demonstrates competitive performance on challenging math benchmarks against strong baselines like DAPO and online DPO.
GRPO is a standard approach to endowing pretrained LLMs with reasoning capabilities. It estimates the advantage of an outcome from a group of $K$ outcomes, and promotes those with positive advantages inside a trust region. Since GRPO discriminates between good and bad outcomes softly, it benefits from additional refinements such as asymmetric clipping and zero-variance data filtering. While effective, these refinements require significant empirical insight and can be challenging to identify. We instead propose an explicit contrastive learning approach. Instead of estimating advantages, we bifurcate $K$ outcomes into positive and negative sets, then maximize the likelihood of positive outcomes. Our approach can be viewed as an online instantiation of (multi-label) noise contrastive estimation for LLM reasoning. We validate our method by demonstrating competitive performance on a suite of challenging math benchmarks against strong baselines such as DAPO and online DPO.