SearcharxivSearch

arXiv subjects

Jiali Wu

Publications and source records attributed to Jiali Wu.

4 recordsLinked to original sources

Towards Understanding What State Space Models Learn About Code

State Space Models (SSMs) have emerged as an efficient alternative to the Transformer architecture. Prior work shows that, when trained under comparable conditions, SSMs can match or surpass Transformers on code understanding tasks. However, their internal mechanisms remain a black box. We present the first systematic analysis of what SSM-based code models learn along with the direct comparison between SSM and Transformer models in this domain. Our analysis shows that SSMs capture syntactic and semantic structure more effectively than Transformers during pretraining but forgets certain relations during fine-tuning on some tasks. To investigate this behavior, we introduce SSM-Interpret, a frequency-domain framework that exposes a spectral shift toward short-range dependencies during fine-tuning. Guided by these findings, we propose architectural modifications that significantly improve the performance of SSM-based code model by upto +6 MRR on NLCodeSearch. This demonstrates that our analysis not only explains model behavior but also leads directly to better designs.

cs.AI

Quantum Meet-in-the-Middle Attacks on Key-Length Extension Constructions

Key-length extension (KLE) techniques provide a general approach to enhancing the security of block ciphers by using longer keys. There are mainly two classes of KLE techniques, cascade encryption and XOR-cascade encryption. This paper presents several quantum meet-in-the-middle (MITM) attacks against two specific KLE constructions. For the two-key triple encryption (2kTE), we propose two quantum MITM attacks under the Q2 model. The first attack, leveraging the quantum claw-finding (QCF) algorithm, achieves a time complexity of $O(2^{2κ/3})$ with $O(2^{2κ/3})$ quantum random access memory (QRAM). The second attack, based on Grover's algorithm, achieves a time complexity of $O(2^{κ/2})$ with $O(2^κ)$ QRAM. The latter complexity is nearly identical to Grover-based brute-force attack on the underlying block cipher, indicating that 2kTE does not enhance security under the Q2 model when sufficient QRAM resources are available. For the 3XOR-cascade encryption (3XCE), we propose a quantum MITM attack applicable to the Q1 model. This attack requires no QRAM and has a time complexity of $O(2^{(κ+n)/2})$ ($κ$ and $n$ are the key length and block length of the underlying block cipher, respectively.), achieving a quadratic speedup over classical MITM attack. Furthermore, we extend the quantum MITM attack to quantum sieve-in-the-middle (SITM) attack, which is applicable for more constructions. We present a general quantum SITM framework for the construction $ELE=E^2\circ L\circ E^1$ and provide specific attack schemes for three different forms of the middle layer $L$. The quantum SITM attack technique can be further applied to a broader range of quantum cryptanalysis scenarios.

cs.CR

The Solar Upper Transition Region Imager (SUTRI) onboard the SATech-01 satellite

The Solar Upper Transition Region Imager (SUTRI) onboard the Space Advanced Technology demonstration satellite (SATech-01), which was launched to a sun-synchronous orbit at a height of 500 km in July 2022, aims to test the on-orbit performance of our newly developed Sc-Si multi-layer reflecting mirror and the 2kx2k EUV CMOS imaging camera and to take full-disk solar images at the Ne VII 46.5 nm spectral line with a filter width of 3 nm. SUTRI employs a Ritchey-Chretien optical system with an aperture of 18 cm. The on-orbit observations show that SUTRI images have a field of view of 41.6'x41.6' and a moderate spatial resolution of 8" without an image stabilization system. The normal cadence of SUTRI images is 30 s and the solar observation time is about 16 hours each day because the earth eclipse time accounts for about 1/3 of SATech-01's orbit period. Approximately 15 GB data is acquired each day and made available online after processing. SUTRI images are valuable as the Ne VII 46.5 nm line is formed at a temperature regime of 0.5 MK in the solar atmosphere, which has rarely been sampled by existing solar imagers. SUTRI observations will establish connections between structures in the lower solar atmosphere and corona, and advance our understanding of various types of solar activity such as flares, filament eruptions, coronal jets and coronal mass ejections.

astro-ph.SR

First-principle prediction of the existence of C64-graphyne and its nitrogen and boron substitutions

By using of the first-principles calculations based on density functional theory, a novel monolayer planar structure named C64-graphyne is predicted. Tetratomic and hexatomic rings, as well as C-C triple bonds exist in this new stable structure with the lattice parameter of 9.291 Å. The carbon hexatomic ring in C64-graphyne contains two quite distinct C-C bonds, which is known as cyclohexatriene. Its electronic band structures show a semiconductor nature with a narrow direct band gap of 0.35 eV. Interestingly, by substituting one nitrogen atom for a carbon atom in the hexatomic ring or a sp hybridized carbon atom on the carbon chain of C64-graphyne, two stable planar structures are obtained. Such doping with nitrogen atom induced metal properties in this monolayer structure. Further investigation shows that the alternating substitution of boron and nitrogen atoms for all carbon atoms in C64-graphyne also induced a new stable structure, the (BN)64 structure. The electronic studies proved its insulator property with the band gap of 4.08 eV.

cond-mat.mtrl-sci