Reinforcement learning from verifiable rewards (RLVR) for mathematical reasoning suffers from a structural blind spot: on "cliff" prompts-those on which every sampled rollout in a group fails-the group-normalized advantage is identically zero, so GRPO produces no gradient on precisely the prompts at the frontier of the model's capability. We introduce LoRA Scaffolded Policy Optimization (LSPO), a sampling-time mechanism that recovers this lost gradient.
arXiv:2607. 27787v1 Announce Type: new Abstract: Reinforcement learning from verifiable rewards (RLVR) for mathematical reasoning suffers from a structural blind spot: on "cliff" prompts-those on which every sampled rollout in a group fails-the group-normalized advantage is identically zero, so GRPO produces no gradient on precisely the prompts at the frontier of the model's capability.
By Ken Ding
Group Adaptive Clipping Policy Optimization (GAPO) is a plug‑in modification to GRPO methods that adapts the importance‑sampling clipping boundary based on rollout advantage. By allowing rollouts with larger learning signals to receive proportionally greater update headroom, GAPO addresses the limitation of fixed clipping that suppresses rare but informative rollouts. Experiments on Qwen and Llama models show that GAPO consistently improves Pass@1 and Pass@k on math reasoning and coding benchmarks where base model pass rates are low.
By Sheng Jia, Xiao Wang, Shiva Prasad Kasiviswanathan, Rein Houthooft
arXiv:2601. 03895v2 Announce Type: replace-cross Abstract: Group Relative Policy Optimization (GRPO) has emerged as a popular algorithm for reinforcement learning with large language models (LLMs).
By Chi Liu, Xin Chen
Group Relative Policy Optimization (GRPO) assigns a magnitude to each rollout based on within‑group reward statistics, rewarding rollouts that reach the correct answer through reasoning. However, the same magnitude can be high for rollouts that reach the answer by guessing, creating a spurious advantage that misleads the policy toward guess‑like behaviors. The paper identifies three scenarios where this occurs—bounded‑answer tasks, open‑answer sets with bounded sub‑cases, and search agents with many paths to the same answer—and proposes SIGNBALANCE, a composition‑free magnitude that preserves the verifier sign, uses a global scale, and restores zero‑mean balance via stop‑gradient per‑class rescaling, matching GRPO on open‑answer math and improving on bounded‑answer math and search agents.
By Jiamian Wang, Samyadeep Basu, Koustava Goswami, Tong Yu, Zhiqiang Tao
arXiv:2607. 26253v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) is bottlenecked by rollout generation, yet many sampled prompts produce saturated groups (all responses correct or all incorrect) whose zero reward variance yields no policy-gradient signal.
By Pixel Nomand, Elena Voss, Marcus Hale, Sofia Reyes
arXiv:2607. 26862v1 Announce Type: new Abstract: Group Relative Policy Optimization (GRPO) has become a standard reinforcement learning method for post-training language models.
By Junoh Park, Junseo Hwang, Wonguk Cho, Taesup Kim
arXiv:2608. 03467v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) com- monly optimizes each correct completion as an independent learning signal.
By Zhe Cao, Miaowen Wen, Fangjiong Chen
The paper critiques the common reinforcement‑learning approach of sampling tool subsets when the full set of tools is enumerable, showing that sampling leads to degraded policy estimates and increased reward sparsity in genomic reasoning tasks. It proposes Full‑Group Policy Optimization (FGPO), which evaluates every tool subset and precomputes rewards in a table, thereby eliminating the need for frozen‑reasoner calls during training. Experiments across five frozen reasoners and three genomic benchmarks demonstrate that FGPO consistently outperforms GRPO, improving average scores by 6.75 points and reducing the number of invoked tools per question.
By Haoyue Liu, Xiaoyu Ma, Ye Chen, Zhichao Wang, Xiaoying Tang
Reinforcement learning over a frozen reasoner has become a common recipe for teaching a policy which external tools to invoke. We show that this recipe becomes structurally mismatched in specialist sc...
The paper identifies a flaw in Group Relative Policy Optimization (GRPO) called the spurious advantage, where the advantage estimator can incorrectly reward rollouts that reach correct answers by guessing rather than reasoning. This issue appears in bounded-answer tasks, open-answer sets with bounded sub-cases, and search agents with many paths to the same answer. The authors propose SIGNBALANCE, a new magnitude estimator that removes composition, maintains verifier sign, applies a global scale, and uses stop‑gradient per‑class rescaling to restore zero‑mean balance, achieving comparable or better performance on math and search benchmarks.
The paper introduces GACA, a critic‑free reinforcement learning estimator that adapts credit assignment granularity based on a step‑level uncertainty proxy. GACA assigns higher weight to fine‑grained signals for steps with above‑average negative log‑likelihood, while relying on episode‑level signals for less uncertain steps, improving task success on ALFWorld and WebShop for 1.5B and 7B language models. The authors provide a risk decomposition, a conditional bound on action‑value variation, and an error‑projection analysis to justify the method’s effectiveness.
By Taoran Liang, Yang Liu, Shang Luo, Yingguang Yang, Rongrong Zhang, Yingzong Min, Yulin Huang, Jianshen Zhang, Yongzhi Qi, Kefu Xu, Congjing Ran, Bin Chong