Elucidating reaction mechanisms hinges on efficiently generating transition states (TSs), products, and complete reaction networks. Recent generative models, such as diffusion models for TS sampling and sequence-based architectures for product generation, offer faster alternatives to quantum-chemistry searches. But diffusion models remain constrained by their stochastic differential equation (SDE) dynamics, which suffer from inefficiency and limited controllability. We show that flow matching, a deterministic ordinary differential (ODE) formulation, can replace SDE-based diffusion for molecular and reaction generation. We introduce MolGEN, a conditional flow-matching framework that learns an optimal transport path to transport Gaussian priors to target chemical distributions. On benchmarks used by TSDiff and OA-ReactDiff, MolGEN surpasses TS geometry accuracy and barrier-height prediction while reducing sampling to sub-second inference. MolGEN also supports open-ended product generation with competitive top-k accuracy and avoids mass/electron-balance violations common to sequence models. In a realistic test on the $γ$-ketohydroperoxide decomposition network, MolGEN yields higher fractions of valid and intended TSs with markedly fewer quantum-chemistry evaluations than string-based baselines. These results demonstrate that deterministic flow matching provides a unified, accurate, and computationally efficient foundation for molecular generative modeling, signaling that flow matching is the future for molecular generation across chemistry.

Chemical structure elucidation is a serious bottleneck in analytical chemistry today. We address the problem of identifying an unknown chemical threat given its mass spectrum and its chemical formula, a task which might take well trained chemists several days to complete. Given a chemical formula, there could be over a million possible candidate structures. We take a data driven approach to rank these structures by using neural networks to predict the presence of substructures given the mass spectrum, and matching these substructures to the candidate structures. Empirically, we evaluate our approach on a data set of chemical agents built for unknown chemical threat identification. We show that our substructure classifiers can attain over 90% micro F1-score, and we can find the correct structure among the top 20 candidates in 88% and 71% of test cases for two compound classes.

This is the second of two papers about MOLGEN.

The selection of what to do next is often the hardest part of problem solving. This selection can be structured by di,'Mguishing decisions about the problem from decisions about the problem solving process. When planning decisions are structured in this way, we find that many of the most important decisionc, are about the planning process itself. This exercise tends to expose a variety of decisons, which are usually made implicitly and sub-optimally in planning programs with rigid control structures. This paper develops a layered approach for meta-planning, that is, for planning about planning. It is part of a course of research which seeks to enhance the power of a problem solver by enabling it to reason about its own reasoning processes.

Problem solvers need more than just the facts and logic of a problem domain. To work effectively, they also aced meta-knowledge, that is, knowledge about how to use the facts. For example, hierarchical planning uses meta-knowledge to distinguish between the important considerations and the details of a problem. This knowledge is added to the facts and logic of a domain and used to focus the generation of inferences. It relieves a hierarchical planner from trying to deal with everything at once.

The selection of what to do next is often the hardest part of resource-limited problem solving. In planning problems, there are typically many goals to be achieved in some order. The goals interact with each other in ways which depend both on the order in which they are achieved and on the particular operators which are used to achieve them. A planning program needs to keep its options open because decisions about one part of a plan are likely to have consequences for another part. This paper describes an approach to planning which integrates and extends two strategies termed the least-commitment and the heuristic strategies.