Metraipolis › archive › dc_885ccd88cf08
Not All Rollouts Are Worth Learning: On Trajectory Valuation for Post-Training Reinforcement Learning — arXiv:2609.35072
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.35072
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:38:47.132Z
- SHA-256 of the archived text
8b088bf49432192626567c9e7d070f420cc4f8f9bc3c0b586fe541c5890846f3- Stable URL of this page
https://metraipolis.com/documento/dc_885ccd88cf08
How to cite
Not All Rollouts Are Worth Learning: On Trajectory Valuation for Post-Training Reinforcement Learning — arXiv:2609.35072 — arXiv (en), https://arxiv.org/abs/2609.35072 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_885ccd88cf08 (Metraipolis archive dc_885ccd88cf08, sha256 8b088bf49432192626567c9e7d070f420cc4f8f9bc3c0b586fe541c5890846f3)
BibTeX
@misc{dc885ccd88cf08,
title = {Not All Rollouts Are Worth Learning: On Trajectory Valuation for Post-Training Reinforcement Learning — arXiv:2609.35072},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.35072},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_885ccd88cf08; sha256 8b088bf49432192626567c9e7d070f420cc4f8f9bc3c0b586fe541c5890846f3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35072. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We consider the problem of trajectory valuation in reinforcement learning: how to identify and mitigate detrimental trajectories during online training. Unlike classification, where data valuation relies on fixed training and validation sets, reinforcement learning involves dynamically generated trajectories without explicit validation signals, making conventional influence-based methods inapplicable. We propose Dynamic Trajectory Valuation (DTV), a simple and efficient framework that estimates trajectory utility at the mini-batch level and filters detrimental trajectories based solely on gradient information. By operating at the optimization level, DTV integrates seamlessly with existing reinforcement learning pipelines with minimal overhead. Extensive experiments across diverse settings, including PPO, GRPO, and DPO, demonstrate that DTV consistently improves performance, enhances data efficiency, and stabilizes optimization. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.35072 — 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 2026-10-07T22:14:28.340Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).