Metraipolis › archive › dc_a42eaa8cdcbe

Two-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLMs — arXiv:2609.03955

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.03955
License (as archived): CC0-1.0
Archived: 2026-10-17T01:40:53.016Z
SHA-256 of the archived text: fc5d1c267f1b723ae4d9de6506886c63a7ad3ec7bb00a83e570de74825ebddd9
Stable URL of this page: https://metraipolis.com/documento/dc_a42eaa8cdcbe

How to cite

Two-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLMs — arXiv:2609.03955 — arXiv (en), https://arxiv.org/abs/2609.03955 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a42eaa8cdcbe (Metraipolis archive dc_a42eaa8cdcbe, sha256 fc5d1c267f1b723ae4d9de6506886c63a7ad3ec7bb00a83e570de74825ebddd9)

BibTeX

@misc{dca42eaa8cdcbe,
  title = {Two-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLMs — arXiv:2609.03955},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.03955},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a42eaa8cdcbe; sha256 fc5d1c267f1b723ae4d9de6506886c63a7ad3ec7bb00a83e570de74825ebddd9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03955. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning (RL) has substantially advanced code generation with large language models (LLMs) through executable feedback. The feedback for coding problems mainly comes from specific test cases, where high-quality test cases are often scarce since they should be both sound and discriminative. We thus turn to study the auto-generation of test cases using the learned model. We find this is naturally an adversarial RL problem: the model is expected to generate effective test cases as counterexamples, depending on the solver's current failure modes. We propose Test Cases Scaling (TCS), a two-stage RL framework for effective test generation. Both stages train a test generator from a rolling policy-aligned buffer: Stage 1 generates tests consistent with the reference solution, and Stage 2 restricts the buffer to current failure modes and learns counterexample tests. Across TACO and LiveCodeBench, TCS improves both pass@1 and inference-time answer selection according to generated tests. We find the learned test generator also enables effective selection among other LLM outputs.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-04 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.03955 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T01:11:43.568Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).