# MagicBeanStalk: Escaping the RLVR Zero Advantage Trap with Recursive Data Synthesis

arXiv: https://arxiv.org/abs/2610.12659 · HTML: https://mlsanity.com/paper/2610.12659/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 30 of its 39 references that are in the map.
- First posted on arXiv 8 Oct 2026.

## Builds on (most cited first, 30 of 30)

- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code](https://mlsanity.com/paper/2403.07974/) (2024)
- [Process Reinforcement through Implicit Rewards](https://mlsanity.com/paper/2502.01456/) (2025)
- [FinanceBench: A New Benchmark for Financial Question Answering](https://mlsanity.com/paper/2311.11944/) (2023)
- [OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction Data](https://mlsanity.com/paper/2410.01560/) (2024)
- [Skywork Open Reasoner 1 Technical Report](https://mlsanity.com/paper/2505.22312/) (2025)
- [Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-Correction](https://mlsanity.com/paper/2508.03613/) (2025)
- [Beyond Benchmarks: MathArena as an Evaluation Platform for Mathematics with LLMs](https://mlsanity.com/paper/2605.00674/) (2026)
- [REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewards](https://mlsanity.com/paper/2505.24760/) (2025)
- [Big-Math: A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models](https://mlsanity.com/paper/2502.17387/) (2025)
- [AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement Learning](https://mlsanity.com/paper/2505.16400/) (2025)
- [Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR](https://mlsanity.com/paper/2508.14029/) (2025)
- [Scaf-GRPO: Scaffolded Group Relative Policy Optimization for Enhancing LLM Reasoning](https://mlsanity.com/paper/2510.19807/) (2025)
- [rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Dataset](https://mlsanity.com/paper/2505.21297/) (2025)
- [RLVE: Scaling Up Reinforcement Learning for Language Models with Adaptive Verifiable Environments](https://mlsanity.com/paper/2511.07317/) (2025)
- [QuestA: Expanding Reasoning Capacity in LLMs via Question Augmentation](https://mlsanity.com/paper/2507.13266/) (2025)
- [SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoning](https://mlsanity.com/paper/2506.08989/) (2025)
- [POPE: Learning to Reason on Hard Problems via Privileged On-Policy Exploration](https://mlsanity.com/paper/2601.18779/) (2026)
- [Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning](https://mlsanity.com/paper/2505.20161/) (2025)
- [CoT-Self-Instruct: Building high-quality synthetic prompts for reasoning and non-reasoning tasks](https://mlsanity.com/paper/2507.23751/) (2025)
- [SCP-116K: A High-Quality Problem-Solution Dataset and a Generalized Pipeline for Automated Extraction in the Higher Education Science Domain](https://mlsanity.com/paper/2501.15587/) (2025)
- [PromptCoT: Synthesizing Olympiad-level Problems for Mathematical Reasoning in Large Language Models](https://mlsanity.com/paper/2503.02324/) (2025)
- [LongPerceptualThoughts: Distilling System-2 Reasoning for System-1 Perception](https://mlsanity.com/paper/2504.15362/) (2025)
- [Golden Goose: A Simple Trick to Synthesize Unlimited RLVR Tasks from Unverifiable Internet Text](https://mlsanity.com/paper/2601.22975/) (2026)
- [SCALER:Synthetic Scalable Adaptive Learning Environment for Reasoning](https://mlsanity.com/paper/2601.04809/) (2026)
- [SECQUE: A Benchmark for Evaluating Real-World Financial Analysis Capabilities](https://mlsanity.com/paper/2504.04596/) (2025)
- [Socratic-MCTS: Test-Time Visual Reasoning by Asking the Right Questions](https://mlsanity.com/paper/2506.08927/) (2025)
- [Reasoning Core: A Scalable RL Environment for LLM Symbolic Reasoning](https://mlsanity.com/paper/2509.18083/) (2025)
- [Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale](https://mlsanity.com/paper/2511.05705/) (2025)
- [DeltaPrompts: Escaping the Zero-Delta Trap in Multimodal Distillation](https://mlsanity.com/paper/2605.15532/) (2026)

## Abstract (arXiv)

Data-centric scaling of RLVR has focused on collecting or generating verifiable training problems. However, corpus size alone does not determine whether the dataset provides a useful learning signal for a target policy. To guide the synthesis of effective RLVR data, we experimentally characterize dataset effectiveness as a function of its problems, the target policy, and the training rollout budget. Under this characterization, we show that a large open-source math corpus pooled from widely used RLVR datasets falls almost entirely into a zero-advantage trap: for a fixed target policy and rollout budget, roughly 88% of its problems are solved in every rollout or failed in every rollout. Filtering concentrates what remains, but it can only select, discarding the hard problems. Effective synthetic problems should therefore require the reasoning operations and compositions required to solve those hard problems. Yet their difficulty must be calibrated, so that some rollouts succeed, some fail, and at least one correct solution fits within the training budget. Prompting a teacher to vary hard problems directly does not achieve this. Guided by these empirical principles, we introduce MagicBeanStalk, a top-down decomposition and bottom-up recursive data-synthesis algorithm that grows seed problems into families of substantively new, policy- and budget-calibrated training problems while preserving the seeds' reasoning blueprints. Across 8 math benchmarks, continual RLVR on MagicBeanStalk problems improves gpt-oss-120b from 61.2 to 69.8 (+8.6) and DeepSeek v4Flash from 72.4 to 77.9 (+5.4) on a 16K-token budget. Training on the full 78K source corpus improves by only +1.1 and its best filtered real data subset by +5.8. In finance, we synthesize over 14K verifiable problems from only 340 solutions and improve over base by 6.9 and 2.4 points on SecQue and FinanceBench, respectively.