arXiv subjects

Hongjun Liu

Publications and source records attributed to Hongjun Liu.

30 records · Page 2Linked to original sources

arXiv2024

DocMath-Eval: Evaluating Math Reasoning Capabilities of LLMs in Understanding Long and Specialized Documents

Recent LLMs have demonstrated remarkable performance in solving exam-like math word problems. However, the degree to which these numerical reasoning skills are effective in real-world scenarios, particularly in expert domains, is still largely unexplored. This paper introduces DocMath-Eval, a comprehensive benchmark specifically designed to evaluate the numerical reasoning capabilities of LLMs in the context of understanding and analyzing specialized documents containing both text and tables. We conduct an extensive evaluation of 48 LLMs with Chain-of-Thought and Program-of-Thought prompting methods, aiming to comprehensively assess the capabilities and limitations of existing LLMs in DocMath-Eval. We found that even the current best-performing system (i.e., GPT-4o) still significantly lags behind human experts in solving complex numerical reasoning problems grounded in long contexts. We believe that DocMath-Eval can serve as a valuable benchmark for evaluating LLMs' capabilities in solving challenging numerical reasoning problems within expert domains.

cs.CL↗

arXiv2024

FinanceMath: Knowledge-Intensive Math Reasoning in Finance Domains

We introduce FinanceMath, a novel benchmark designed to evaluate LLMs' capabilities in solving knowledge-intensive math reasoning problems. Compared to prior works, this study features three core advancements. First, FinanceMath includes 1,200 problems with a hybrid of textual and tabular content. These problems require college-level knowledge in the finance domain for effective resolution. Second, we provide expert-annotated, detailed solution references in Python program format, ensuring a high-quality benchmark for LLM assessment. We also construct a finance-domain knowledge bank and investigate various knowledge integration strategies. Finally, we evaluate a wide spectrum of 44 LLMs with both Chain-of-Thought and Program-of-Thought prompting methods. Our experimental results reveal that the current best-performing system (i.e., GPT-4o) achieves only 60.9% accuracy using CoT prompting, leaving substantial room for improvement. Moreover, while augmenting LLMs with external knowledge can improve model performance (e.g., from 47.5% to 54.5% for Gemini-1.5-Pro), their accuracy remains significantly lower than the estimated human expert performance of 92%. We believe that FinanceMath can advance future research in the area of domain-specific knowledge retrieval and integration, particularly within the context of solving reasoning-intensive tasks.

cs.CL↗

arXiv2022

Cryptanalysis and designing chaos-based irreversible and parallel key expansion module over Galois field

From the security criteria of irreversibility, parallelizability and independence, we cryptanalyzed the key expansion modules of candidate block ciphers of AES, the results revealed that there exist some weaknesses inside, which may be explored by the attacker. Hence, we designed a more secure key expansion module that the round-key can satisfy three criteria above. First, we constructed a non-degenerate 2D chaotic map (2D-πeCM) with ergodicity in phase space and sufficient large chaotic range. Then based on 2D-πeCM and polynomial multiplication over Galois field, we designed an irreversible key expansion module, which could transform the initial key of arbitrary length to desired number of independent round keys in parallel. Security and statistical results demonstrated the flexible and effectiveness of the proposed key expansion module.

cs.CR↗

arXiv2022

Complete and robust energy conversion by sum frequency generation based on Invariant Engineering

We propose an analytical method to achieve complete energy conversion in sum frequency generation based on Lewis-Riesenfeld invariants theory. This technique, derived from a two-level atom transition in quantum mechanics, is more efficient and robust than conventional methods. In our scheme, the quasi-adiabatic single control parameter model is established, and the value of single control parameter is selected to make the initial eigenstate perfectly converted to the final eigenstate we need. Corresponds to the nonlinear frequency conversion process, the nonlinear crystal structure is designed with the inverse engineering of optimal control theory, which is robust against the perturbations in the coupling coefficient and phase mismatch, including pump intensity and crystal polarization period variations, and results in almost 100% conversion efficiency at any crystal length. It is demonstrated that the frequency conversion can be achieved in the wavelength range of 2.6 μm -3.6 μm with a spectral bandwidth of the conversion efficiency over 50% approaching to 400 nm when the crystal length L=1 mm.

quant-ph↗

arXiv2022

Scattered Image Reconstruction at Near-infrared Based on Spatial Modulation Instability

We present a method of near-infrared image reconstruction based on spatial modulation instability in a photorefractive strontium barium niobate crystal. The conditions that lead to the formation of modulation instability at near-infrared are discussed depending on the theory of modulation instability gain. Experimental results of scattered image reconstruction at the 1064 nm wavelength show the maximum cross-correlation coefficient and cross-correlation gain are 0.57 and 2.09 respectively. This method is expected to be an aid for near-infrared imaging technologies.

physics.optics↗

arXiv2021

Cryptanalyze and design strong S-Box using 2D chaotic map and apply to irreversible key expansion

Cryptanalysis result of key expansion algorithms in AES and SM4 revealed that, (1) there exist weaknesses in their S-Boxes, and (2) the round key expansion algorithm is reversible, i.e., the initial key can be recovered from any round key, which may be an exploitable weakness by attacker. To solve these problems, first we constructed a non-degenerate 2D exponential hyper chaotic map (2D-ECM), derived the recursion formula to calculate the number of S-Boxes that satisfied three conditions, and designed a strong S-Box construction algorithm without weakness. Then based on 2D-ECM and S-Box, we designed an irreversible key expansion algorithm, to transform the initial key into independent round keys, to make the initial key can not be recovered from any round key. Security and statistical analysis demonstrated the flexible and effectiveness of the proposed irreversible key expansion algorithm.

cs.CR↗

arXiv2020

Exciton interaction induced spin splitting in MoS$_2$ monolayer

By pumping nonresonantly a MoS$_2$ monolayer at $13$ K under a circularly polarized cw laser, we observe exciton energy redshifts that break the degeneracy between B excitons with opposite spin. The energy splitting increases monotonically with the laser power reaching as much as $18$ meV, while it diminishes with the temperature. The phenomenon can be explained theoretically by considering simultaneously the bandgap renormalization which gives rise to the redshift and exciton-exciton Coulomb exchange interaction which is responsible for the spin-dependent splitting. Our results offer a simple scheme to control the valley degree of freedom in MoS$_2$ monolayer and provide an accessible method in investigating many-body exciton exciton interaction in such materials.

cond-mat.mes-hall↗

arXiv2017

Inversion Domain Boundary Induced Stacking and Bandstructure Diversity in Bilayer MoSe2

Interlayer rotation and stacking were recently demonstrated as effective strategies for tuning physical properties of various two-dimensional materials. The latter strategy was mostly realized in hetero-structures with continuously varied stacking orders, which obscure the revelation of the intrinsic role of a certain stacking order in its physical properties. Here, we introduce inversion-domain-boundaries into molecular-beam-epitaxy grown MoSe2 homo-bilayers, which induce unusual-fractional lattice translations to their surrounding domains, accounting for the observed diversity of large-area and uniform stacking sequences. Unusual low-symmetric stacking orders were observed using transmission electron microscopy and detailed geometries were identified by density functional theory. A linear relation was also revealed between interlayer distance and stacking energy. These stacking sequences yield various energy alignments between the valence states at Γ and K, showing stacking dependent bandgaps and valence band tail states in the measured scanning tunneling spectroscopy. These results may benefit the design of two-dimensional multilayers with manipulable stacking orders.

cond-mat.mtrl-sci↗

arXiv2015

Observation of intervalley quantum interference in epitaxial monolayer WSe2

Monolayer (ML) transition metal dichalcogenides (TMDs) have been attracting great research attentions lately for their extraordinary properties, in particular the exotic spin-valley coupled electronic structures that promise future spintronic and valleytronic applications1-3. The energy bands of ML TMDs have well separated valleys that constitute effectively an extra internal degree of freedom for low energy carriers3-12. The large spin-orbit coupling in the TMDs makes the spin index locked to the valley index, which has some interesting consequences such as the magnetoelectric effects in 2H bilayers13. A direct experimental characterization of the spin-valley coupled electronic structure can be of great interests for both fundamental physics and device applications. In this work, we report the first experimental observation of the quasi-particle interference (QPI) patterns in ML WSe2 using low-temperature (LT) scanning tunneling microscopy/spectroscopy (STM/S). We observe intervalley quantum interference involving the Q-valleys in the conduction band due to spin-conserved scattering processes, while spin-flip intervalley scattering is absent. This experiment establishes unequivocally the presence of spin-valley coupling and affirms the large spin-splitting at the Q valleys. Importantly, the inefficient spin-flip intervalley scattering implies long valley and spin lifetime in ML WSe2, which represents a key figure of merit for valley-spintronic applications.

cond-mat.mtrl-sci↗

arXiv2015

Line and Point Defects in MoSe2 Bilayer Studied by Scanning Tunneling Microscopy and Spectroscopy

Bilayer (BL) MoSe2 films grown by molecular-beam epitaxy (MBE) are studied by scanning tunneling microscopy and spectroscopy (STM/S). Similar to monolayer (ML) films, networks of inversion domain boundary (DB) defects are observed both in the top and bottom layers of BL MoSe2, and often they are seen spatially correlated such that one is on top of the other. There are also isolated ones in the bottom layer without companion in the top-layer and are detected by STM/S through quantum tunneling of the defect states through the barrier of the MoSe2 ML. Comparing the DB states in BL MoSe2 with that of ML film reveals some common features as well as differences. Quantum confinement of the defect states is indicated. Point defects in BL MoSe2 are also observed by STM/S, where ionization of the donor defect by the tip-induced electric field is evidenced. These results are of great fundamental interests as well as practical relevance of devices made of MoSe2 ultrathin layers.

cond-mat.mtrl-sci↗

arXiv2014

Dense network of one-dimensional mid-gap metallic modes in monolayer MoSe2 and their spatial undulations

We report the observation of a dense triangular network of one-dimensional (1D) metallic modes in a continuous and uniform monolayer of MoSe2 grown by molecular-beam epitaxy. High-resolution transmission electron microscopy and scanning tunneling microscopy and spectroscopy (STM/STS) studies show these 1D modes are mid-gap states at inversion domain boundaries. STM/STS measurements further reveal intensity undulations of the metallic modes, presumably arising from the superlattice potentials due to moire pattern and the quantum confinement effect. A dense network of the metallic modes with high density of states is of great potential for heterocatalysis applications. The interconnection of such mid-gap 1D conducting channels may also imply new transport behaviors distinct from the 2D bulk.

cond-mat.mtrl-sci↗

arXiv2012

Windowed Carbon Nanotubes for Efficient CO2 Removal from Natural Gas

We demonstrate from molecular dynamics simulations that windowed carbon nanotubes can efficiently separate CO2 from the CO2/CH4 mixture, resembling polymeric hollow fibers for gas separation. Three CO2/CH4 mixtures with 30%, 50% and 80% CO2 are investigated as a function of applied pressure from 80 to 180 bar. In all simulated conditions, only CO2 permeation is observed; CH4 is completely rejected by the nitrogen-functionalized windows or pores on the nanotube wall in the accessible timescale, while maintaining a fast diffusion rate along the tube. The estimated time-dependent CO2 permeance ranges from 107 to 105 GPU (gas permeation unit), compared with ~100 GPU for typical polymeric membranes. CO2/CH4 selectivity is estimated to be ~108 from the difference in free-energy barriers of permeation. This work suggests that a windowed carbon nanotube can be used as a highly efficient medium, configurable in hollow-fiber-like modules, for removing CO2 from natural gas.

cond-mat.mtrl-sci↗