arXiv subjects

Qituan Zhang

Publications and source records attributed to Qituan Zhang.

3 recordsLinked to original sources

arXiv2026

Rank-one 4d $\mathcal N=3$ SCFTs: Schur index, VOA modules, and modularity

We study the representation theory of the vertex operator algebras (VOAs) associated with rank-one 4d $\mathcal{N} = 3$ superconformal field theories. For the $\mathbb{Z}_3$ S-fold theory, whose VOA $\mathcal{W}_{\mathbb{Z}_3}$ has central charge $c_{\mathrm{2d}} = -15$, we use the $\mathcal{N} = 1$ Lagrangian description to obtain the unflavored Schur index in terms of Dedekind eta functions, while Wilson-loop indices yield the unflavored non-vacuum characters. These characters all solve a modular linear differential equation (MLDE) whose solution space also contains a logarithmic character. Combining flavored MLDEs from null states with Zhu's associative algebra and a free-field realization, we study four highest-weight modules of $\mathcal{W}_{\mathbb{Z}_3}$ and their flavored characters in closed form. A parallel analysis applies to the $\mathcal{N} = 3$ theories obtained by gauging a discrete $\mathbb{Z}_n$ flavor subgroup of $\mathcal{N} = 4$ $U(1)$ and $SU(2)$ super-Yang--Mills, for which we also obtain closed-form Schur indices and a new free-field realization of the VOA of the $\mathbb{Z}_4$ quotient.

hep-th↗

arXiv2026

Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to $5.49\times$ end-to-end speedup under the Transformers backend and up to $5.8\times$ throughput speedup under SGLang serving.

cs.CL↗

arXiv2024

Superconductivity of Bulk Abnormal Magic-stoichiometric Na3Cl Salt Crystals at Normal Pressure

The identification of new materials with superconducting properties is the pursuit in the realm of superconductivity research. Here, excitedly, we show that the simplest salt daily used can be made a superconductor at normal pressure only by adjusting its stoichiometry of Na and Cl as Na3Cl at normal pressure based on first-principles calculations. This bulk stable abnormal Na-Cl stoichiometric crystal of 3:1, the first 'magic' ratio, includes metallic (Na) atoms in the core as well as hybridization of ionic and metallic bonding, facilitating the electron-phonon-coupling for superconductivity with a critical temperature Tc of 0.13 K. The flat bands and van Hove singularities near the Fermi level produce large densities of states, similar to H3S and LaH10, which is beneficial for the emergence of superconductivity. The crystal composed of with abnormal Na-Cl magic stoichiometry is a precisely tunable, purely sodium and chloride-based, three-dimensional bulk superconductor, which is therefore an ideal material for designing and understanding abnormal stoichiometric crystals. The methodology of constructing this bulk abnormal crystal may be general to almost all elements, which could lead to insights into the physics of other conventional superconductors and even high-critical-temperature superconductors.

cond-mat.supr-con↗