{"ID":6537514,"CreatedAt":"2026-07-14T02:54:43.516908796Z","UpdatedAt":"2026-07-15T03:28:55.185153975Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.11506","arxiv_id":"2607.11506","title":"SCOPE-RL: Optimizing Reasoning Paths Before and After Success","abstract":"Reinforcement learning with verifiable rewards (RLVR) optimizes LLMs using sparse verifiable final-answer rewards. This sparse anchor reliably verifies whether a trajectory succeeds but provides no direct feedback on the reasoning path that produced it. Before success, prerequisite progress on hard problems receives no reward signal; after success, outcome rewards cannot distinguish well-organized correct trajectories from redundant or locally flawed ones. We introduce SCOPE-RL (Scaffolded Chain Optimization with Process Efficiency), a two-stage framework that densifies this anchor while retaining the GRPO update: Adaptive Scaffolded RL adds prefix-decomposed verifiable rewards on answer-hidden sub-question chains before success, and Quality-Aware Process RL applies correctness-gated process-shape rewards to refine correct trajectories after success. An expert-validated Step-Quality Evaluation Protocol evaluates useful-step density, error localization, and token efficiency beyond final-answer accuracy. On Qwen3-8B-Instruct trained on DAPO-Math and Big-Math, SCOPE-RL improves average accuracy by up to 11.2 pp and reduces reasoning tokens by up to 27.1% over outcome-only GRPO; the gains hold under GSPO and on Qwen3-0.6B-Instruct, indicating that reward-signal densification is complementary to policy-update-level RLVR advances. Code and data are available at https://github.com/tokencraft-lab/SCOPE-RL.","short_abstract":"Reinforcement learning with verifiable rewards (RLVR) optimizes LLMs using sparse verifiable final-answer rewards. This sparse anchor reliably verifies whether a trajectory succeeds but provides no direct feedback on the reasoning path that produced it. Before success, prerequisite progress on hard problems receives no...","url_abs":"https://arxiv.org/abs/2607.11506","url_pdf":"https://arxiv.org/pdf/2607.11506v1","authors":"[\"Xiaojian Liu\",\"Han Xu\",\"Jianqiang Xia\",\"Zhixuan Li\",\"Ke Xu\",\"Yiwei Dai\",\"Xinran Chen\",\"Changwo Wu\",\"Yuchen Li\"]","published":"2026-07-13T12:56:28Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CL\"]","methods":"[\"Reinforcement Learning\",\"Large Language Model\",\"Generative Adversarial Network\"]","has_code":false,"code_links":[{"ID":614209,"CreatedAt":"2026-07-14T02:54:43.516908796Z","UpdatedAt":"2026-07-14T02:54:43.516908796Z","DeletedAt":null,"paper_id":6537514,"paper_url":"https://arxiv.org/abs/2607.11506","paper_title":"SCOPE-RL: Optimizing Reasoning Paths Before and After Success","repo_url":"https://github.com/tokencraft-lab/SCOPE-RL","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}