Climate change and anthropogenic activities have exacerbated hypoxic conditions in the global ocean, posing a serious threat to marine ecosystems. Quantifying changes in dissolved oxygen levels is crucial for understanding the impact of these factors on marine life and earth sustainability. However, dissolved oxygen observations are severely sparse, limiting comprehensive analysis. Here we present Jingwei, a 4D spatiotemporal graph transfer learning model that reconstructs ocean oxygen levels globally over the past six decades on a 1◦ x 1◦ grid, covering depths of 0-5500 meters. Jingwei simultaneously leverages intra-profile knowledge transfer using a pattern bank from simulations and captures inter-profile correlations through zoning-varying message passing among observations. It significantly reduces reconstruction error by 27.26% compared to CMIP6, demonstrates high consistency with cruise surveys, and provides satisfactory visual quality. Furthermore, Jingwei produces interpretable results, effectively identifying vertical profile patterns and distinguishing fine-grained spatial distributions. Jingwei provides cartography and quantitative analysis of oxygen minimum zones (OMZ) evolution since 1960. We foresee Jingwei revolutionizing observation-based ocean modeling and deepening our understanding of the breathless ocean. Alongside, we have released an open-source online platform (https://jingwei.acemap.info/ https://jingwei.acemap.info/), providing data visualization, resource sharing and ongoing updates.
Bin Lu, Luyu Han, Ze Zhao et al.· Proceedings of the 32nd ACM...· 0 citations
Reinforcement learning has become a promising paradigm for improving large language model (LLM) agents on long-horizon search tasks, where the agent must make a sequence of intermediate decisions before receiving a final outcome. However, existing methods still face a key limitation: the rollout budget is often allocated without explicitly assessing the utility of intermediate states. As a result, substantial computation may be spent on low-value states, even though different branches can vary drastically in their informativeness. In this paper, we propose Information Gain-based Rollout Policy Optimization (IGRPO), a policy optimization framework that treats intermediate-state informativeness as the organizing principle of rollout collection. Specifically, IGRPO performs budget-aware tree-structured rollouts by allocating expansion budget according to node-level informativeness, so that more informative branches are expanded more frequently while unpromising branches are progressively suppressed. We further demonstrate that the information gain-based rollout induces an explicit limiting teacher distribution over trajectories, which naturally yields a clear policy optimization target, thereby unifying adaptive tree-structured exploration with principled policy learning under a single framework. Experiments on seven challenging search-augmented QA benchmarks demonstrate that IGRPO consistently outperforms strong baselines under the same rollout budget constraints, validating the effectiveness of leveraging the induced teacher distribution to guide policy optimization for long-horizon search agents.
Yijun Zhang, Fan Xu, Jiaxin Ding et al.· 2 citations
The first oceanic 4D sparse observation reconstruction dataset, named OceanVerse, is presented, providing a novel large-scale dataset that meets the MNAR (Missing Not at Random) condition, supporting more effective model comparison, generalization evaluation and potential advancement of scientific reconstruction architectures.
Bin Lu, Jingjing Shen, Ze Zhao et al.· Proceedings of the 32nd ACM...· 0 citations