Scaling the Scaling Logic: Agentic Meta-Synthesis of Logic Reasoning

Reinforcement Learning from Verifiable Rewards (RLVR) is bottlenecked by data: existing synthesis pipelines rely on expert-written code or fixed templates, confining growth to instance-level perturbations. We shift the evolvable unit from problem instances to task-family specifications. SSLogic is an agentic meta-synthesis framework in which LLM agents iteratively author and refine executable Generator-Validator pairs inside a closed Generate-Validate-Refine loop, producing families with new rules and difficulty gradients rather than parameter variations of old ones. A Multi-Gate Validation Protocol, multi-strategy consensus plus Adversarial Blind Review, where independent agents solve each instance by writing and executing code, filters ill-posed tasks before they enter training. Starting from 400 seed families, two evolution rounds yield 953 families and 21,389 verifiable instances. Three converging comparisons (step-matched, token-matched, and size-controlled on external Enigmata data) consistently show higher training utility of evolved data, with gains of SynLogic +5.2, AIME25 +3.0, and BBH +5.5 on Enigmata. Fine-grained KORBench evaluation reveals selective improvements in logic (+13.2%) and operation (+9.6%), linking structural evolution to downstream gains. Code is available at https://github.com/AdAstraAbyssoque/Scaling-the-Scaling-Logic.

This paper studies how to create many reliable practice problems for improving the reasoning ability of AI systems. Instead of manually writing every problem or only changing the numbers in existing templates, our method asks AI agents to design whole families of new problems together with programs that can check whether the answers are correct. Several independent checks are used to filter out unclear or unsolvable problems before they are used for training. Starting from 400 initial problem families, the method produces 953 families and more than 21,000 verified problem instances. Experiments show that training on these generated problems improves performance on several reasoning benchmarks, including logic, math, and broader reasoning tasks. The goal is to make high-quality reasoning training data easier to scale while keeping the answers automatically checkable.