EvoSyn: Generalizable Evolutionary Data Synthesis for Verifiable Learning

TL;DR

EvoSyn introduces a universal, evolutionary data synthesis framework that generates verifiable, diverse training data for language models, improving their performance across multiple tasks without relying on domain-specific heuristics.

Contribution

The paper presents a novel, task-agnostic data synthesis method that jointly creates problems, solutions, and verification artifacts, enabling robust, generalizable training data generation.

Findings

Significant performance improvements on LiveCodeBench and AgentBench-OS.

Effective generalization across different domains and tasks.

Reliable, principled synthesis of verifiable training data.

Abstract

Reliable verifiable data has become a key driver of capability gains in modern language models, enabling stable reinforcement learning with verifiable rewards and effective distillation that transfers competence across math, coding, and agentic tasks. Yet constructing generalizable synthetic verifiable data remains difficult due to hallucination-prone generation, and weak or trivial verification artifacts that fail to separate strong from weak solutions. Existing approaches often rely on task-specific heuristics or post-hoc filters that do not transfer across domains and lack a principled, universal evaluator of verifiability. In this work, we introduce an evolutionary, task-agnostic, strategy-guided, executably-checkable data synthesis framework that, from minimal seed supervision, jointly synthesizes problems, diverse candidate solutions, and verification artifacts, and iteratively discovers strategies via a consistency-based evaluator that enforces agreement between human-annotated and strategy-induced checks. This pipeline upgrades filtering into principled synthesis: it reliably assembles coherent, verifiable training instances and generalizes without domain-specific rules. Our experiments demonstrate the effectiveness of the proposed approach under both RLVR and model distillation training paradigms. The results show that training with our synthesized data yields significant improvements on both the LiveCodeBench and AgentBench-OS tasks, highlighting the robust generalization of our framework.