Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMs

TL;DR

Ring-lite is a scalable, efficient large language model that combines MoE and reinforcement learning, achieving strong reasoning performance with fewer activated parameters and improved training stability.

Contribution

The paper introduces C3PO, a novel RL training algorithm for MoE models, and a two-stage training paradigm for multi-domain data integration, enhancing stability and efficiency.

Findings

Matches SOTA small-scale reasoning models on benchmarks

Activates only one-third of parameters compared to similar models

Demonstrates improved training stability with C3PO

Abstract

We present Ring-lite, a Mixture-of-Experts (MoE)-based large language model optimized via reinforcement learning (RL) to achieve efficient and robust reasoning capabilities. Built upon the publicly available Ling-lite model, a 16.8 billion parameter model with 2.75 billion activated parameters, our approach matches the performance of state-of-the-art (SOTA) small-scale reasoning models on challenging benchmarks (e.g., AIME, LiveCodeBench, GPQA-Diamond) while activating only one-third of the parameters required by comparable models. To accomplish this, we introduce a joint training pipeline integrating distillation with RL, revealing undocumented challenges in MoE RL training. First, we identify optimization instability during RL training, and we propose Constrained Contextual Computation Policy Optimization(C3PO), a novel approach that enhances training stability and improves computational throughput via algorithm-system co-design methodology. Second, we empirically demonstrate that selecting distillation checkpoints based on entropy loss for RL training, rather than validation metrics, yields superior performance-efficiency trade-offs in subsequent RL training. Finally, we develop a two-stage training paradigm to harmonize multi-domain data integration, addressing domain conflicts that arise in training with mixed dataset. We will release the model, dataset, and code.