Metraipolis › archive › dc_862e00cc7dc5
Hints, Critics, and Teachers: Prior Injection for Sparse-Reward RL in Vision-Language Math Reasoning — arXiv:2608.21811
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.21811
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T07:38:51.853Z
- SHA-256 of the archived text
e9bf8afa8c8b077b5152e26126c7518fe207b2da9ac49ec559b93b699b4595ad- Stable URL of this page
https://metraipolis.com/documento/dc_862e00cc7dc5
How to cite
Hints, Critics, and Teachers: Prior Injection for Sparse-Reward RL in Vision-Language Math Reasoning — arXiv:2608.21811 — arXiv (en), https://arxiv.org/abs/2608.21811 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_862e00cc7dc5 (Metraipolis archive dc_862e00cc7dc5, sha256 e9bf8afa8c8b077b5152e26126c7518fe207b2da9ac49ec559b93b699b4595ad)
BibTeX
@misc{dc862e00cc7dc5,
title = {Hints, Critics, and Teachers: Prior Injection for Sparse-Reward RL in Vision-Language Math Reasoning — arXiv:2608.21811},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.21811},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_862e00cc7dc5; sha256 e9bf8afa8c8b077b5152e26126c7518fe207b2da9ac49ec559b93b699b4595ad},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.21811. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement learning for vision-language math reasoning starves under sparse reward: on a pool of 20,830 visual-math problems where Qwen2-VL-2B answers 3.6% of rollouts correctly, 85-97% of GRPO rollout groups are entirely wrong and contribute zero gradient. We train eleven methods under identical conditions in this regime, each injecting a different prior: text (reference-solution hints), distribution (on-policy distillation from a 7B teacher), and value (a value-pretrained critic with an MSE or HL-Gauss categorical loss). A prior helps exactly when it is delivered: the six arms whose prior effectively reaches the policy separate with no overlap from the remaining five -- the no-prior baseline and four arms whose prior is teacher-capped, gated away, or lost to a mis-parameterized critic -- both on the pooled in-domain metric and on cross-domain transfer (DynaMath). The central finding, however, concerns evaluation: one slice of the in-domain pool -- long used as this project's general-distribution check -- anti-correlates with genuine cross-domain transfer (Spearman rho = -0.74, n = 11 arms, permutation p = 0.011), while the hardest in-domain slice predicts it closely (rho = +0.89, p < 0.001). We attribute the inversion to a near-chance multiple-choice subset that rewards models for not having changed; read through it, the best cross-domain method looked mediocre and the worst looked like the champion. Among the methods, hint-guided exploration -- not UFT's auxiliary loss -- drives hint gains, and replacing the critic's MSE loss with HL-Gauss cross-entropy is worth +14.4 points in-domain. All accuracies are blind-judged, with paired exact tests. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-25 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.21811 — 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 2026-10-18T07:37:52.670Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).