Metraipolis › archive › dc_0fc8bdd1e666

Information-Gain Rewards over Diversity-Pruned Tests: GT-Anchored Verifier Co-Training for Reliable Code Generation — arXiv:2609.21208

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.21208
License (as archived)
CC0-1.0
Archived
2026-10-10T01:01:43.114Z
SHA-256 of the archived text
77fd1d3b2b4676c1c6189cb8d7b97ec4af08007e0b9c2583e4cd2d3b109a5937
Stable URL of this page
https://metraipolis.com/documento/dc_0fc8bdd1e666

How to cite

Information-Gain Rewards over Diversity-Pruned Tests: GT-Anchored Verifier Co-Training for Reliable Code Generation — arXiv:2609.21208 — arXiv (en), https://arxiv.org/abs/2609.21208 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0fc8bdd1e666 (Metraipolis archive dc_0fc8bdd1e666, sha256 77fd1d3b2b4676c1c6189cb8d7b97ec4af08007e0b9c2583e4cd2d3b109a5937)
BibTeX
@misc{dc0fc8bdd1e666,
  title = {Information-Gain Rewards over Diversity-Pruned Tests: GT-Anchored Verifier Co-Training for Reliable Code Generation — arXiv:2609.21208},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.21208},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0fc8bdd1e666; sha256 77fd1d3b2b4676c1c6189cb8d7b97ec4af08007e0b9c2583e4cd2d3b109a5937},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.21208. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Self-play methods that co-train a single language model as both coder and test author promise to move code-generation RL beyond fixed test suites, but they suffer from two coupled pathologies: permissiveness collapse, where pass-rate rewards are maximised by trivial, non-discriminative tests, and concentration bias, where i.i.d. sampled tests cluster on modal inputs and inflate estimator variance. We introduce CoVer (Co-trained Coder and Verifier), a single-policy GRPO framework that addresses both failure modes. First, an information-gain (IG) reward scores each self-generated test by the mutual information between its pass/fail vector and a graded, ground-truth-anchored correctness signal y [0, 1] m, gated by the sign of their covariance so that only positively discriminative tests receive reward. Second, a three-stage diversity-aware selection step prunes a candidate pool to a behaviourally non-redundant suite (invalidity, input-string, execution-profile filtering), raising the effective sample size of the IG estimator at fixed execution budget. On five benchmarks (LiveBench, MBPP, LiveCodeBench, CodeContests, Code-Forces), CoVer raises one-shot pass@1 by +5.8 points at 7B and +7.1 points at 14B over the Qwen2.5-Instruct backbone, and achieves the highest macro-average among all compared methods at both scales. As a drop-in backbone inside the CodeT ranking pipeline, CoVer-7B adds +3.5 points, demonstrating the dual benefit of co-training for both generation and selection.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-21 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.21208 — 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 2026-10-10T00:23:42.437Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-21?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).