Components
No mentions found
This entity hasn't been tracked yet, or Iris is still building its knowledge base.
Related Articles from SNS
Diving into Kronecker Adapters: Component Design Matters
arXiv:2602.01267v2 Announce Type: replace Abstract: Kronecker adapters have emerged as a promising approach for fine-tuning large-scale models, enabling high-rank updates through tunable component structures. However, existing work largely treats the component structure as a fixed or heuristic design choice, leaving the dimensions and number of Kronecker components underexplored. In this paper, we identify component structure as a key factor governing the capacity of Kronecker adapters.
Enhancing Blind Source Separation with Dissociative Principal Component Analysis
arXiv:2411.12321v2 Announce Type: replace Abstract: Principal component analysis (PCA) and its sparse variants (sPCA) are widely used as a precursor to independent component analysis (ICA) for blind source separation (BSS). However, sPCA typically relies on a deflation strategy that extracts components sequentially and imposes orthogonality between them. When the underlying sources overlap, this discards the cross component structure that ICA depends on, degrading separation.
Poking Around in the Dark: Why a Shared Understanding of Components Matters
arXiv:2606.02442v1 Announce Type: new Abstract: By listing the components included in an application, Software Bills of Materials (SBOMs) are intended to support the timely identification of vulnerable components and ensure the security of the software supply chain. However, we question the underlying assumption that there is agreement on the components to be listed in an SBOM and that current technology is sufficient to secure the software supply chain. First, we propose a ground-up...
Component Ablation for Efficient Hybrid Language Model Architectures: Performance, Resilience, and Compression Implications
arXiv:2603.22473v2 Announce Type: replace Abstract: Hybrid language models combine softmax attention with linear-time sequence mechanisms such as state-space or linear-attention layers, but the functional contribution of each component type remains insufficiently characterized. We study component-level ablation in two sub-1B hybrid language models, Qwen3.5-0.8B and Falcon-H1-0.5B, using likelihood-based evaluation, downstream benchmarks, layer-wise interventions, random controls, and...
Multi-component Causal Tracing in Large Language Models
Announce Type: new Abstract: Causal tracing systematically intervenes on a large language model's (LLM's) internal representations to uncover and quantify the causal pathways linking specific inputs or computations to specific metrics of interest, quantifying the LLM's behavior. Building on previous single-component or single-layer studies, this paper presents a unified framework for causally tracing multiple components simultaneously. This framework systematically identifies the subsets of...
Fiber optic components enable high-performance 2-µm fiber lasers
Fiber optic components enable high-performance 2-µm fiber lasers Lisa Lock Scientific Editor Robert Egan Associate Editor Laser systems operating in the 2-micrometer wavelength range open diverse opportunities in medical technology, agriculture, and plastics processing. In the Eurostars project DECOMP, Laser Zentrum Hannover e.V. (LZH) has developed novel fiber optic components that overcome previous technical barriers.
Two-component exciton condensates in an electron–hole bilayer
Abstract Macroscopic quantum coherence emerges when bosons condense into a Bose–Einstein condensate (BEC)1,2,3,4,5. Excitons are a long-sought solid-state route to high-temperature BECs with strong interactions, electrical tunability and potentially multicomponent spinor order, but conclusive evidence for equilibrium condensation has remained elusive. Here we report evidence for two-component exciton BECs in MoSe2/hBN/WSe2 electron–hole bilayers6,7,8,9 by probing the spin–valley...
Vision Transformer Finetuning Benefits from Non-Smooth Components
arXiv:2602.06883v3 Announce Type: replace Abstract: The smoothness of the transformer architecture has been extensively studied in the context of generalization, training stability, and adversarial robustness. However, its role in transfer learning remains poorly understood. In this paper, we analyze the ability of vision transformer components to adapt their outputs to changes in inputs, or, in other words, their \emph{plasticity}.
Quaternion Dirac--Coulomb--Breit Integral Transformation for Relativistic Four-Component Correlated Electronic Structure Theory
Announce Type: new Abstract: High-accuracy correlated four-component relativistic electronic structure methods are typically formulated in terms of integrals over molecular orbital (MO). Consequently, an efficient and scalable strategy is required to deal with the complexity of transforming relativistic two-electron integrals from the atomic orbital (AO) to the MO basis. The transformation bottleneck is particularly acute for approaches that include Breit interaction integrals, whose...
Ukraine’s Flamingo missiles hit Russian factory producing key drone components
KYIV — Several Ukrainian Flamingo FP-5 missiles travelled over 1,000 kilometers into Russia and struck a factory that produces key components for deadly drones, Ukrainian President Volodymyr Zelenskyy said Wednesday. The cruise missiles targeted the Progress factory in Cheboksary, capital of Russia’s Chuvash Republic. The facility produces the Kometa antennas that enable Russian drones and missiles to bypass Ukrainian air defenses during the Kremlin’s massive strikes on the country.