Search

Stack Overflow is a prominent Q and A forum, supporting developers in seeking suitable resources on programming-related matters. Having high-quality question titles is an effective means to attract developers' attention. Unfortunately, this is often underestimated, leaving room for improvement. Research has been conducted, predominantly leveraging pre-trained models to generate titles from code snippets and problem descriptions. Yet, getting high-quality titles is still a challenging task, attributed to both the quality of the input data (e.g., containing noise and ambiguity) and inherent constraints in sequence generation models. In this paper, we present FILLER as a solution to generating Stack Overflow post titles using a fine-tuned language model with self-improvement and post ranking. Our study focuses on enhancing pre-trained language models for generating titles for Stack Overflow posts, employing a training and subsequent fine-tuning paradigm for these models. To this end, we integrate the model's predictions into the training process, enabling it to learn from its errors, thereby lessening the effects of exposure bias. Moreover, we apply a post-ranking method to produce a variety of sample candidates, subsequently selecting the most suitable one. To evaluate FILLER, we perform experiments using benchmark datasets, and the empirical findings indicate that our model provides high-quality recommendations. Moreover, it significantly outperforms all the baselines, including Code2Que, SOTitle, CCBERT, M3NSCT5, and GPT3.5-turbo. A user study also shows that FILLER provides more relevant titles, with respect to SOTitle and GPT3.5-turbo., Comment: The paper has been per-reviewed and accepted for publication to the International Symposium on Empirical Software Engineering and Measurement (ESEM 2024)

In the last decades, scene text recognition has gained worldwide attention from both the academic community and actual users due to its importance in a wide range of applications. Despite achievements in optical character recognition, scene text recognition remains challenging due to inherent problems such as distortions or irregular layout. Most of the existing approaches mainly leverage recurrence or convolution-based neural networks. However, while recurrent neural networks (RNNs) usually suffer from slow training speed due to sequential computation and encounter problems as vanishing gradient or bottleneck, CNN endures a trade-off between complexity and performance. In this paper, we introduce SAFL, a self-attention-based neural network model with the focal loss for scene text recognition, to overcome the limitation of the existing approaches. The use of focal loss instead of negative log-likelihood helps the model focus more on low-frequency samples training. Moreover, to deal with the distortions and irregular texts, we exploit Spatial TransformerNetwork (STN) to rectify text before passing to the recognition network. We perform experiments to compare the performance of the proposed model with seven benchmarks. The numerical results show that our model achieves the best performance., Comment: Accepted to ICMLA 2020

A topological insulator doped with random magnetic impurities is studied. The system is modelled by the Kane-Mele model with a random spin exchange between conduction electrons and magnetic dopants. The dynamical mean field theory for disordered systems is used to investigate the electron dynamics. The magnetic long-range order and the topological invariant are calculated within the mean field theory. They reveal a rich phase diagram, where different magnetic long-range orders such as antiferromagnetic or ferromagnetic one can exist in the metallic or insulating phases, depending on electron and magnetic impurity fillings. It is found that insulator only occurs at electron half filling, quarter filling and when electron filling is equal to magnetic impurity filling. However, non-trivial topology is observed only in half-filling antiferromagnetic insulator and quarter-filling ferromagnetic insulator. At electron half filling, the spin Hall conductance is quantized and it is robust against magnetic doping, while at electron quarter filling, magnetic dopants drive the ferromagnetic topological insulator to ferromagnetic metal. The quantum anomalous Hall effect is observed only at electron quarter filling and dense magnetic doping.

Metal-insulator transitions in a three-component Falicov-Kimball model are investigated within the Kotliar-Ruckenstein slave boson mean-field approach. The model describes a mixture of two interacting fermion atom species loaded into an optical lattice at ultralow temperature. One species is two-component atoms, which can hop in the optical lattice, and the other is single-component atoms, which are localized. Different correlation-driven metal-insulator transitions are observed depending on the atom filling conditions and local interactions. These metal-insulator transitions are classified by the band renormalization factors and the double occupancies of the atom species. The filling conditions and the critical value of the local interactions for these metal-insulator transitions are also analytically established. The obtained results not only are in good agreement with the dynamical mean-field theory for the three-component Falicov-Kimball model but also clarify the nature and properties of the metal-insulator transitions in a simple physics picture.

We study the charge ordering behavior under a magnetic field $H$ in the extended Hubbard model within the coherent potential approximation. At quarter filling, for small $H$ we find that the relative variation of critical temperature is quadratic with the coefficient ${\alpha}$ smaller than the one for conventional spin-Peierls systems. For intermediate field, a melting of the charge ordering on decreasing temperature under fixed $H$ at various band filling is found., Comment: 9 pages, 4 figures