Beyond Surface: Probing LLaMA Across Scales and Layers

This paper presents an in-depth analysis of Large Language Models (LLMs), focusing on LLaMA, a prominent open-source foundational model in natural language processing. Instead of assessing LLaMA through its generative output, we design multiple-choice tasks to probe its intrinsic understanding in high-order tasks such as reasoning and computation. We examine the model horizontally, comparing different sizes, and vertically, assessing different layers. We unveil several key and uncommon findings based on the designed probing tasks: (1) Horizontally, enlarging model sizes almost could not automatically impart additional knowledge or computational prowess. Instead, it can enhance reasoning abilities, especially in math problem solving, and helps reduce hallucinations, but only beyond certain size thresholds; (2) In vertical analysis, the lower layers of LLaMA lack substantial arithmetic and factual knowledge, showcasing logical thinking, multilingual and recognitive abilities, with top layers housing most computational power and real-world knowledge.

Large Language Models are Diverse Role-Players for Summarization Evaluation

Text summarization has a wide range of applications in many scenarios. The evaluation of the quality of the generated text is a complex problem. A big challenge to language evaluation is that there is a clear divergence between existing metrics and human evaluation. For example, the quality of a document summary can be measured by human annotators from both objective aspects, such as grammatical and semantic correctness, as well as subjective dimensions, such as comprehensiveness, succinctness, and interestingness. Most of the automatic evaluation methods like BLUE/ROUGE may be not able to capture the above dimensions well. In this paper, we propose a new evaluation framework based on LLMs, which provides a comprehensive evaluation framework by comparing generated text and reference text from both objective and subjective aspects. First, we propose to model objective and subjective dimensions of generated text based on roleplayers prompting mechanism. Furthermore, we introduce a context-based prompting mechanism that is able to generate dynamic roleplayer profiles based on input context. Finally, we design a multi-roleplayer prompting technology based on batch prompting to integrate multiple evaluation results into evaluation results. Experimental results on two real datasets for summarization show that our model is highly competitive and has a very high consistency with human annotators.

Multi-level Contrastive Learning for Cross-lingual Spoken Language Understanding

Although spoken language understanding (SLU) has achieved great success in high-resource languages, such as English, it remains challenging in low-resource languages mainly due to the lack of high quality training data. The recent multilingual code-switching approach samples some words in an input utterance and replaces them by expressions in some other languages of the same meaning. The multilingual code-switching approach achieves better alignments of representations across languages in zero-shot cross-lingual SLU. Surprisingly, all existing multilingual code-switching methods disregard the inherent semantic structure in SLU, i.e., most utterances contain one or more slots, and each slot consists of one or more words. In this paper, we propose to exploit the "utterance-slot-word" structure of SLU and systematically model this structure by a multi-level contrastive learning framework at the utterance, slot, and word levels. We develop novel code-switching schemes to generate hard negative examples for contrastive learning at all levels. Furthermore, we develop a label-aware joint model to leverage label semantics for cross-lingual knowledge transfer. Our experimental results show that our proposed methods significantly improve the performance compared with the strong baselines on two zero-shot cross-lingual SLU benchmark datasets.

Reinforced Iterative Knowledge Distillation for Cross-Lingual Named Entity Recognition

Named entity recognition (NER) is a fundamental component in many applications, such as Web Search and Voice Assistants. Although deep neural networks greatly improve the performance of NER, due to the requirement of large amounts of training data, deep neural networks can hardly scale out to many languages in an industry setting. To tackle this challenge, cross-lingual NER transfers knowledge from a rich-resource language to languages with low resources through pre-trained multilingual language models. Instead of using training data in target languages, cross-lingual NER has to rely on only training data in source languages, and optionally adds the translated training data derived from source languages. However, the existing cross-lingual NER methods do not make good use of rich unlabeled data in target languages, which is relatively easy to collect in industry applications. To address the opportunities and challenges, in this paper we describe our novel practice in Microsoft to leverage such large amounts of unlabeled data in target languages in real production settings. To effectively extract weak supervision signals from the unlabeled data, we develop a novel approach based on the ideas of semi-supervised learning and reinforcement learning. The empirical study on three benchmark data sets verifies that our approach establishes the new state-of-the-art performance with clear edges. Now, the NER techniques reported in this paper are on their way to become a fundamental component for Web ranking, Entity Pane, Answers Triggering, and Question Answering in the Microsoft Bing search engine. Moreover, our techniques will also serve as part of the Spoken Language Understanding module for a commercial voice assistant. We plan to open source the code of the prototype framework after deployment.

CalibreNet: Calibration Networks for Multilingual Sequence Labeling

Lack of training data in low-resource languages presents huge challenges to sequence labeling tasks such as named entity recognition (NER) and machine reading comprehension (MRC). One major obstacle is the errors on the boundary of predicted answers. To tackle this problem, we propose CalibreNet, which predicts answers in two steps. In the first step, any existing sequence labeling method can be adopted as a base model to generate an initial answer. In the second step, CalibreNet refines the boundary of the initial answer. To tackle the challenge of lack of training data in low-resource languages, we dedicatedly develop a novel unsupervised phrase boundary recovery pre-training task to enhance the multilingual boundary detection capability of CalibreNet. Experiments on two cross-lingual benchmark datasets show that the proposed approach achieves SOTA results on zero-shot cross-lingual NER and MRC tasks.

A Multi-level Neural Network for Implicit Causality Detection in Web Texts

Mining causality from text is a complex and crucial natural language understanding task. Most of the early attempts at its solution can group into two categories: 1) utilizing co-occurrence frequency and world knowledge for causality detection; 2) extracting cause-effect pairs by using connectives and syntax patterns directly. However, because causality has various linguistic expressions, the noisy data and ignoring implicit expressions problems induced by these methods cannot be avoided. In this paper, we present a neural causality detection model, namely Multi-level Causality Detection Network (MCDN), to address this problem. Specifically, we adopt multi-head self-attention to acquire semantic feature at word level and integrate a novel Relation Network to infer causality at segment level. To the best of our knowledge, in touch with the causality tasks, this is the first time that the Relation Network is applied. The experimental results on the AltLex dataset, demonstrate that: a) MCDN is highly effective for the ambiguous and implicit causality inference; b) comparing with the regular text classification task, causality detection requires stronger inference capability; c) the proposed approach achieved state-of-the-art performance.