BORG enhanced

We have engineered BORG to accommodate a wide array of physical models, a necessity for scaling BORG inference to cosmological volume. These models are enhanced by Machine Learning techniques, developed in conjunction with AFM WG. Notably, we have broken new ground in non-linear N-body inference by integrating a neural network emulator of the displacement field (Jamieson et al., 2023), (Doeser et al., 2024) with the BORG inference machine. We have introduced a novel BORG-Velocity model (Boruah et al., 2022), (Prideaux-Ghee et al., 2023), potentially offering the most precise handling of redshift space distortions in galaxy surveys to date. We have developed new galaxy bias model capable of going beyond the simple remappings that we were using so far, and be resistant to model misspecifications (Ding et al., 2024). We have also investigated innovative methods for computing modified gravity evolution, enabling its incorporation into our differentiable simulator. Our team has developed new differentiable galaxy weak lensing software, rigorously validated against analytical calculations. This validation process, utilized by ILI-WG, has facilitated the creation of new weak lensing data compression for ILI. We are also developing new metrics to assess the quality of peculiar velocity reconstruction with regard to distance data (Stiskalek et al., 2026). These metrics were used to assess the accuracy of the new Manticore-Main inference (McAlpine et al., 2025) for which we found orders of magnitude of improvements over all available distance tracers. Exceeding our initial objectives, we have pioneered a new technology for differentiable Zoom simulation in BORG (Wempe et al., 2024). Furthermore we provided key technology to the ILI WG for running fast simulation at scales over the cosmological volume with BORG-PM in the GoBig pipeline.