Consensus Group Relative Policy Optimization (C‑GRPO) is a new training method that distills Minimum Bayes Risk (MBR) decoding into a group‑relative objective, enabling text generation models to approximate MBR performance without the costly inference‑time sampling and scoring. C‑GRPO only needs a utility function and policy samples, avoiding the need for gold references or curated preference data. Experiments on WMT 2024 machine translation and XSum summarization show that C‑GRPO matches MBR decoding quality while reducing inference overhead and outperforming other reference‑free baselines.
arXiv:2602.14028v2 Announce Type: replace Abstract: While Group Relative Policy Optimization (GRPO) offers a powerful framework for LLM post-training, its effectiveness in open-ended domains like Mac...
arXiv:2606. 04807v1 Announce Type: new Abstract: Mitigating social bias in Large Language Models (LLMs) presents a distinct alignment challenge: unlike verifiable tasks, bias lacks a single ground truth, creating a high-variance, subjective reward landscape.
Mitigating social bias in Large Language Models (LLMs) presents a distinct alignment challenge: unlike verifiable tasks, bias lacks a single ground truth, creating a high-variance, subjective reward landscape. Previous preference-based fine-tuning methods have major trade-offs: Direct Preference Optimization (DPO) is limited by the lack of exploration inherent in offline training, while Proximal Policy Optimization (PPO) can lead to training instability due to potentially unreliable critic estimates.
The paper introduces Personalized Group Relative Policy Optimization (P‑GRPO), a new alignment framework for large language models that separates advantage estimation from immediate batch statistics. By normalizing advantages using preference‑group‑specific reward histories instead of the concurrent generation group, P‑GRPO maintains contrastive signals for distinct user preferences. Experiments across various tasks show that P‑GRPO converges faster and yields higher rewards than standard GRPO, improving alignment with heterogeneous human preferences while preserving general capabilities.
arXiv:2510. 05342v2 Announce Type: replace-cross Abstract: Direct Preference Optimization (DPO) has emerged as a simple and effective method for aligning large language models.
arXiv:2602. 19327v3 Announce Type: replace Abstract: A significant portion of recent research on Large Language Model (LLM) alignment focuses on developing new policy optimization methods based on Group Relative Policy Optimization (GRPO).
GUPO: Gradient Uncertainty-aware Policy Optimization for Post-Training Large Language Models proposes a new method for fine‑tuning LLMs after training. The approach models each group gradient as a random variable, estimates its probability distribution, and uses Dirichlet‑based gradient uncertainty to weight each group’s contribution during policy updates. Experiments on multiple benchmarks show that this uncertainty‑aware aggregation improves the effectiveness of post‑training policy optimization.
arXiv:2606. 28707v1 Announce Type: new Abstract: Critic-free reinforcement learning with verifiable rewards (RLVR), exemplified by Group Relative Policy Optimization (GRPO), avoids training a value function (critic) and reduces memory and compute overhead relative to critic-based PPO pipelines for aligning large language models.
arXiv:2608. 03119v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) improves LLM reasoning but typically relies on ground-truth (GT) answers, limiting scalability.
ConsensusBench is a new dataset that supplies rule‑based process‑level signals for large language model reasoning. It identifies key intermediate conclusions—called Consensus Nodes—by filtering correct trajectories and clustering semantically equivalent statements. By incorporating a process reward derived from these nodes into GRPO‑style reinforcement learning, the authors create ConsensusPR, which reduces reward sparsity and improves performance on benchmarks such as AIME, GSM8K, and MATH‑500.
The paper introduces Best‑Practice Critic Optimization (BPCO), a stable and efficient recipe for training a critic in reinforcement learning for large language models. BPCO combines DPPO, bounded value predictions, Monte Carlo targets, unnormalized policy advantages, and length‑adaptive advantage estimation, allowing the critic to be conditioned on hidden reward information. Experiments on mathematical reasoning tasks with models from 1.5B to 30B parameters show that BPCO consistently outperforms a strong critic‑based baseline and matches or exceeds group‑based methods while sampling only one response per prompt.
The paper introduces Token Importance-Aware Policy Optimization (TIAO), a reinforcement learning approach that improves text summarization by weighting token importance based on token dependency. TIAO reweights a trajectory’s advantage according to the overall dependencies of core tokens, addressing the limitation of previous methods that treat all tokens equally. Experiments demonstrate that a 7B foundation model enhanced with TIAO achieves performance comparable to GPT‑4 and GPT‑5‑nano on real‑world datasets.