Search

The most recent pointwise Large Language Model (LLM) rankers have achieved remarkable ranking results. However, these rankers are hindered by two major drawbacks: (1) they fail to follow a standardized comparison guidance during the ranking process, and (2) they struggle with comprehensive considerations when dealing with complicated passages. To address these shortcomings, we propose to build a ranker that generates ranking scores based on a set of criteria from various perspectives. These criteria are intended to direct each perspective in providing a distinct yet synergistic evaluation. Our research, which examines eight datasets from the BEIR benchmark demonstrates that incorporating this multi-perspective criteria ensemble approach markedly enhanced the performance of pointwise LLM rankers.

Active learning (AL), which aims to construct an effective training set by iteratively curating the most formative unlabeled data for annotation, has been widely used in low-resource tasks. Most active learning techniques in classification rely on the model's uncertainty or disagreement to choose unlabeled data, suffering from the problem of over-confidence in superficial patterns and a lack of exploration. Inspired by the cognitive processes in which humans deduce and predict through causal information, we take an initial attempt towards integrating rationales into AL and propose a novel Explainable Active Learning framework (XAL) for low-resource text classification, which aims to encourage classifiers to justify their inferences and delve into unlabeled data for which they cannot provide reasonable explanations. Specifically, besides using a pre-trained bi-directional encoder for classification, we employ a pre-trained uni-directional decoder to generate and score the explanation. We further facilitate the alignment of the model with human reasoning preference through a proposed ranking loss. During the selection of unlabeled data, the predicted uncertainty of the encoder and the explanation score of the decoder complement each other as the final metric to acquire informative data. Extensive experiments on six datasets show that XAL achieves consistent improvement over 9 strong baselines. Analysis indicates that the proposed method can generate corresponding explanations for its predictions., Comment: Accepted by NAACL 2024

Magicity, or shell closure, plays an important role in our understanding of complex nuclear phenomena. In this work, we employ one of the state-of-the-art density functional theories, the deformed relativistic Hartree-Bogoliubov theory in continuum (DRHBc) with the density functional PC-PK1, to investigate the evolution of the $N=20,28,50$ shell closures in the $ 20 \leqslant Z \leqslant 30$ region. We show how these three conventional shell closures evolve from the proton drip line to the neutron drip line by studying the charge radii, two-neutron separation energies, two-neutron gaps, quadrupole deformations, and single-particle levels. In particular, we find that in the $ 21 \leqslant Z \leqslant 27$ region, the $N=50$ shell closure disappears or becomes quenched, mainly due to the deformation effects. Similarly, both experimental data and theoretical predictions indicate that the $N=28$ shell closure disappears in the Mn isotopic chain, also predominantly due to the deformation effects. The DRHBc theory predicts the existence of the $N=20$ shell closure in the Ca, Sc, and Ti isotopic chains, but the existing data for the Ti isotopes suggests the contrary, and therefore more investigations are needed., Comment: 9 pages, 9 figures

Non-coplanar Intensity-Modulated Radiation Therapy (IMRT) goes a step further by orienting the gantry carrying the radiation beam and the patient couch in a non-coplanar manner to accurately target the cancer region and better avoid organs-at-risk. The use of a non-coplanar treatment trajectory significantly enhances the degree of freedom and flexibility but increases drastically the complexity of the optimization. In inverse planning optimization the dose contribution for all potential beam directions is usually pre-calculates and pre-loads into the Treatment Planning System (TPS). The size the dose matrix becomes more critical when moving from coplanar IMRT to non-coplanar IMRT since the number of beams increases drastically. A solution would be to calculate "on-the-fly" the dose contribution to each new candidate beam during optimization. This is only possible if a dose calculation engine is fast enough to be used online during optimization iterations, which is not the case in standard method. Therefore, in this work we propose an IMRT optimization scheme using deep learning based dose engine to compute the dose matrix on-line. The proposed deep learning approach will be combined into a simulated-annealing-based optimization method for non-coplanar IMRT. Since the dose engine will compute the dose contribution on-line during the optimization, the final main optimization method requires to keep in memory a very lightweight dose matrix. The proposed method was compared with clinical data showing a good agreement considering dosimetry of the treatment plans. The main advantage of the proposed method was the reduction of the memory storage from 9GB to 10MB during the optimization process., Comment: 12 pages, 6 figures, 2 tables

Cross-domain sentiment analysis aims to predict the sentiment of texts in the target domain using the model trained on the source domain to cope with the scarcity of labeled data. Previous studies are mostly cross-entropy-based methods for the task, which suffer from instability and poor generalization. In this paper, we explore contrastive learning on the cross-domain sentiment analysis task. We propose a modified contrastive objective with in-batch negative samples so that the sentence representations from the same class will be pushed close while those from the different classes become further apart in the latent space. Experiments on two widely used datasets show that our model can achieve state-of-the-art performance in both cross-domain and multi-domain sentiment analysis tasks. Meanwhile, visualizations demonstrate the effectiveness of transferring knowledge learned in the source domain to the target domain and the adversarial test verifies the robustness of our model.

Automated event detection from news corpora is a crucial task towards mining fast-evolving structured knowledge. As real-world events have different granularities, from the top-level themes to key events and then to event mentions corresponding to concrete actions, there are generally two lines of research: (1) theme detection identifies from a news corpus major themes (e.g., "2019 Hong Kong Protests" vs. "2020 U.S. Presidential Election") that have very distinct semantics; and (2) action extraction extracts from one document mention-level actions (e.g., "the police hit the left arm of the protester") that are too fine-grained for comprehending the event. In this paper, we propose a new task, key event detection at the intermediate level, aiming to detect from a news corpus key events (e.g., "HK Airport Protest on Aug. 12-14"), each happening at a particular time/location and focusing on the same topic. This task can bridge event understanding and structuring and is inherently challenging because of the thematic and temporal closeness of key events and the scarcity of labeled data due to the fast-evolving nature of news articles. To address these challenges, we develop an unsupervised key event detection framework, EvMine, that (1) extracts temporally frequent peak phrases using a novel ttf-itf score, (2) merges peak phrases into event-indicative feature sets by detecting communities from our designed peak phrase graph that captures document co-occurrences, semantic similarities, and temporal closeness signals, and (3) iteratively retrieves documents related to each key event by training a classifier with automatically generated pseudo labels from the event-indicative feature sets and refining the detected key events using the retrieved documents. Extensive experiments and case studies show EvMine outperforms all the baseline methods and its ablations on two real-world news corpora., Comment: Accepted to KDD 2022 Research Track

A standard-form Wadati-Konno-Ichikawa(WKI) type integrable hierarchy is derived from a corresponding matrix spectral problem associated with the Lie algebra sl(2, R). Each equation in the resulting hierarchy has a bi-Hamiltonian structure furnished by the trace identity. Then, the higher grading affine algebraic construction of some special cases is proposed. We also show that eneralized short pulse equation arises naturally from the negative WKI flow.