Welcome to Gist.Science
We read the latest papers from arXiv, bioRxiv, and medRxiv and produce easy-to-understand explanations, key takeaways, and technical summaries — in ten languages.
Imagine you are watching a river where two streams of water are flowing side-by-side at different speeds. One is fast, one is slow. Usually, where they meet, they start to swirl and mix, creating beautiful, chaotic eddies.…
This paper introduces ReasonXL, a large-scale multilingual reasoning corpus, and demonstrates that a two-stage fine-tuning and reinforcement learning pipeline can successfully adapt LLMs to reason in non-English languages without sacrificing performance, while revealing that language identity is determined by early model layers and adaptation is concentrated in upper layers.
SOLARIS is a novel framework that enables real-time inference for complex foundation models in recommendation systems by proactively precomputing and asynchronously offloading user-item embeddings, thereby achieving significant revenue gains at Meta's scale without compromising serving latency.
This study demonstrates that the impact of memory length on cooperation in LLM-based social swarms is fundamentally determined by model-specific characteristics, such as alignment and personality interpretation, where Gemini-2.0-Flash agents suppress cooperation with longer memories due to negative sentiment accumulation, while Gemma-3 agents promote it.
This paper experimentally demonstrates a novel two-dimensional quantum-path interference in high-harmonic generation driven by orthogonally polarized, highly-bichromatic fields, where distinct modulation patterns in odd and even harmonics reveal the dynamic symmetry of multiple quantum orbits and offer a new pathway for attosecond electron spectroscopy.
Using state-of-the-art cosmological data including DESI DR2, the authors find that a coupled dark energy model with fermionic cold dark matter is viable with a coupling parameter peak at , effectively excluding the no-coupling scenario at 95% confidence while demonstrating robustness across different potentials and supernova samples.
This paper resolves Huang's problem by constructing a unitary weak quasi-Hopf algebra framework via quantum gauge groups to establish a complete identification between the analytic braided tensor structure on affine vertex operator algebra modules and the Huang-Lepowsky structure for Lie types A, B, C, D, and G₂, thereby unifying high-dimensional algebraic QFT with low-dimensional cases and solving a decades-old problem posed by Doplicher and Roberts.
This article reviews the construction of a chaos-assisted low-dimensional holographic correspondence between the SYK model and Jackiw-Teitelboim gravity for non-experts, demonstrating how quantum chaos bridges bulk and boundary physics and necessitates string theory elements to resolve fine-grained quantum scales.
This paper demonstrates that a late-time transition in the standardized absolute magnitude of Type Ia supernovae at approximately 20 Mpc significantly improves the fit to Pantheon+ data and systematically increases the inferred Hubble constant by about 2%, while leaving other cosmological parameters like matter density and dark energy equation of state largely unaffected.
This paper proposes a goal-oriented safe active learning framework for model predictive control that recursively updates Bayesian recurrent neural network parameters through alternating exploration and goal-reaching phases, ensuring safety, finite-time exploration termination, and near-optimal performance while adapting to system dynamics online.
This paper proposes an adaptive sparse group lasso penalized quantile regression method that utilizes a dual ADMM algorithm to achieve simultaneous within- and between-group sparsity with proven global convergence and superior computational efficiency.
This paper demonstrates that using nonconvex penalty functions in sparse regularisation improves the quantification of structural uncertainty in chemical reaction network inference by better covering plausible network structures compared to traditional lasso methods, thereby enabling a hierarchical representation of ambiguities to guide future experimental design.