FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale

Runyuan He, Qiuyang Mang, Shang Zhou, Kaiyuan Liu, Hanchen Li, Huanzhi Mao, Qizheng Zhang, Zerui Li, Bo Peng, Lufeng Cheng, Tianfu Fu, Yichuan Wang, Wenhao Chai, Jingbo Shang, Alex Dimakis, Joseph E. Gonzalez, Alvin Cheung

arXiv:2605.14445 · 2026-05-15 공개 · arXiv · PDF

benchmarking competitive-programming llm-coding frontier-smith idea-divergence long-horizon-coding test-case-generation verifier-generation

Abstract

Many real-world coding challenges are open-ended and admit no known optimal solution. Yet, recent progress in LLM coding has focused on well-defined tasks such as feature implementation, bug fixing, and competitive programming. Open-ended coding remains a weak spot for LLMs, largely because open-ended training problems are scarce and expensive to construct. Our goal is to synthesize open-ended coding problems at scale to train stronger LLM coders. We introduce FrontierSmith, an automated system for iteratively evolving open-ended problems from existing closed-ended coding tasks. Starting from competitive programming problems, FrontierSmith generates candidate open-ended variants by changing the problems'goals, restricting outputs, and generalizing inputs. It then uses a quantitative idea divergence metric to select problems that elicit genuinely diverse approaches from different solvers. Agents then generate test cases and verifiers for the surviving candidates. On two open-ended coding benchmarks, training on our synthesized data yields substantial gains over the base models: Qwen3.5-9B improves by +8.82 score on FrontierCS and +306.36 (Elo-rating-based performance) on ALE-bench; Qwen3.5-27B improves by +12.12 and +309.12, respectively. The synthesized problems also make agents take more turns and use more tokens, similar to human-curated ones, suggesting that closed-ended seeds can be a practical starting point for long-horizon coding data.

한국어 요약

📋 한 줄 요약

**[코딩 LLM / 데이터 합성]** 닫힌 코딩 문제로부터 다양성을 명시적으로 보장한 열린 코딩 문제를 대규모로 합성해 LLM 코더의 약점을 메우는 시스템 FrontierSmith 제안.

🎯 핵심 기여도

💡 핵심 아이디어

열린 코딩 문제는 "어떻게 다양한 접근이 동시에 정답이 될 수 있는가"를 측정해야 하며, 닫힌 시드 문제에서 출발해도 다양성을 정량적으로 필터링하면 사람이 만든 문제와 닮은 long-horizon 학습 데이터를 얻을 수 있다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: open-ended 코딩 데이터의 희소성을 자동 합성으로 해결하는 실용 경로 제시, 닫힌 시드도 long-horizon 학습 출발점이 될 수 있음을 입증. **한계**: 시드 분포가 결국 경쟁 프로그래밍에 편중, 산업 실무 도메인으로의 일반화는 후속 과제.

🚀 실용적 활용