This review critically evaluates Mg-mediated routes for producing non-ferrous metals and related materials, including magnesiothermic extraction and MgCl2-assisted electrochemical processing. The discussion first clarifies the thermodynamic basis and metallurgical significance of Mg before analyzing the reaction pathways, kinetics, and transport limitations in direct solid-state reduction and gas-phase Mg infiltration. Mg-mediated extraction of Ti, Zr, rare earth elements, Nb, Ta, and Si is subsequently examined with emphasis on the coupling of reduction behavior with separation, purification, morphology control, and by-product management. The review also assesses engineering barriers, including Mg vapor containment, reactor corrosion, MgO separation, and the energy penalty of Mg regeneration. Electrochemical processing in molten halides is discussed with particular attention to MgCl2 as a chlorinating agent, an electroactive melt component, and a functional additive that improves process feasibility and efficiency. Finally, future directions are outlined for closed-loop Mg metallurgy and hybrid thermochemical-electrochemical extraction, focusing on electrolysis hardware, Mg recovery, and heat integration. By linking magnesiothermic reduction with molten-halide electrochemistry, this review positions Mg as both an effective reductant and a recyclable redox carrier for low-carbon non-ferrous extractive metallurgy.
The nominal divide between p- and d-electron systems often obscures a deep underlying unity in condensed matter physics. This review elucidates the orbital homology between the p and t2g orbital manifolds, establishing the correspondence that extends from minimal model Hamiltonians to the complex behaviors of real quantum materials. We demonstrate that despite their distinct atomic origins, these orbitals host nearly identical hopping physics and spin–orbit coupling, formalized through an effective l=1 angular momentum algebra for the t2g case. This equivalence allows one to transpose physical intuition and theoretical models developed for p-orbital systems directly onto the more complex t2g materials, and vice versa. We showcase how this paradigm provides a unified understanding of emergent phenomena, including non-trivial band topology, itinerant ferromagnetism, and unconventional superconductivity, across a wide range of platforms, from transition metal compounds, two-dimensional oxide heterostructures, and iron-based superconductors, to p-orbital ultracold gases. Ultimately, this p-t2g homology serves not only as a tool for interpretation but also as a robust design principle for engineering novel quantum states.
Vision-Language-Action (VLA) models typically bridge the gap between perceptual and action spaces by pre-training a large-scale Vision-Language Model (VLM) on robotic data. While this approach greatly enhances performance, it also incurs significant training costs. In this paper, we investigate how to effectively bridge vision-language (VL) representations to action (A). We introduce VLA-Adapter, a novel paradigm designed to reduce the reliance of VLA models on large-scale VLMs and extensive pre-training. To this end, we first systematically analyze the effectiveness of various VL conditions and present key findings on which conditions are essential for bridging perception and action spaces. Based on these insights, we propose a lightweight Policy module with Bridge Attention, which autonomously injects the optimal condition into the action space. In this way, our method achieves high performance using only a 0.5B-parameter backbone, without any robotic data pre-training. Extensive experiments on both simulated and real-world robotic benchmarks show that VLA-Adapter not only achieves state-of-the-art level performance, but also offers the fast inference speed reported to date. Furthermore, thanks to the proposed advanced bridging paradigm, VLA-Adapter enables the training of a powerful VLA model on a single consumer-grade GPU, greatly lowering the barrier to deploying VLA model.
Vision-language-action (VLA) models have recently shown strong potential in enabling robots to follow language instructions and execute precise actions. However, most VLAs are built upon vision-language models pretrained solely on 2D data, which lack accurate spatial awareness and hinder their ability to operate in the 3D physical world. Existing solutions attempt to incorporate explicit 3D sensor inputs such as depth maps or point clouds, but these approaches face challenges due to sensor noise, hardware heterogeneity, and incomplete depth coverage in existing datasets. Alternative methods that estimate 3D cues from 2D images also suffer from the limited performance of depth estimators. We propose Spatial Forcing (SF), a simple yet effective alignment strategy that implicitly forces VLA models to develop spatial comprehension capabilities without relying on explicit 3D inputs or depth estimators. SF aligns intermediate visual embeddings of VLAs with geometric representations produced by pretrained 3D foundation models. By enforcing alignment at intermediate layers, SF guides VLAs to encode richer spatial representations that enhance action precision. Extensive experiments in simulation and real-world environments demonstrate that SF achieves state-of-the-art results, surpassing both 2D- and 3D-based VLAs. SF further accelerates training by up to 3.8× and improves data efficiency across diverse robotic tasks.
Context. Characterizing the masses, radii, and compositions of small planets orbiting M dwarfs is key to understanding their formation and identifying the best targets for atmospheric follow-up with facilities such as JWST. Methods. We refined the photometry of the TOI-4342 system using TESS and LCOGT data, and characterized the host stars with NIRPS and ESPRESSO spectroscopy. High-precision ESPRESSO radial velocities (RVs) allowed us to constrain the planetary masses and investigate their potential compositions. Results. The TOI-4336 A system is composed of a sub-Neptune with a period of 16.34 days, a radius of 2.14 +/- 0.08 R-circle plus, and a mass of 3.33 +/- 0.36 M-circle plus, along with an inner super-Earth on a 7.59-day orbit with a radius of 1.25 +/- 0.07 R-circle plus and a mass of 1.55 +/- 0.13 M-circle plus. The TOI-4342 system hosts two sub-Neptunes of similar sizes (2.33 +/- 0.09 R-circle plus and 2.35 +/- 0.09 R-circle plus), with periods of 5.54 and 10.69 days. Their masses are measured to be 7.3 +/- 1.3 M-circle plus and 4.8 +/- 1.4 M-circle plus, respectively. The RVs also reveal a planet candidate around TOI-4342, most likely non-transiting, with a period of 47.5 days and a minimum mass of 17.8 +/- 3.0 M-circle plus. Conclusions. With precise radii and masses, we derived bulk densities and explored possible compositions. The TOI-4336 A subNeptune and super-Earth have densities of 1.87 +/- 0.30 and 4.35 +/- 0.79 g cm(-3), while the two similar-sized sub-Neptunes in TOI-4342 show distinct densities of 3.18 +/- 0.67 and 2.01 +/- 0.63 g cm(-3). Using an inference model, we find that TOI-4336 A b, TOI-4342 b, and TOI-4342 c have an atmosphere mass fraction (AMF) of similar to 3.7%, similar to 1.8%, and similar to 2.9%, respectively, while the super-Earth TOI-4336 A c could contain similar to 2% of water or have a core-to-mass fraction (CMF) of similar to 31%. All four planets are excellent targets for future atmospheric characterization with JWST, and their multi-planet nature makes them especially interesting for comparative planetology. Notably, TOI-4336 A b stands out as one of the best known targets in its size and temperature regime, with a transmission spectroscopy metric (TSM) of 138, comparable to benchmark planets such as K2-18 b and LHS 1140 b. Its inner sibling, TOI-4336 A c, may also be of interest for emission spectroscopy and exploring the "cosmic shoreline", similarly to the Rocky Worlds DDT JWST program.
