Hybrid quantum-classical classifiers promise to positively impact critical
aspects of natural language processing tasks, particularly
classification-related ones. Among the possibilities currently investigated,
quantum transfer learning, i.e., using a quantum circuit for fine-tuning
pre-trained classical models for a specific task, is attracting significant
attention as a potential platform for proving quantum advantage.
This work shows potential advantages, both in terms of performance and
expressiveness, of quantum transfer learning algorithms trained on embedding
vectors extracted from a large language model to perform classification on a
classical Linguistics task: acceptability judgments. Acceptability judgment is
the ability to determine whether a sentence is considered natural and
well-formed by a native speaker. The approach has been tested on sentences
extracted from ItaCoLa, a corpus that collects Italian sentences labeled with
their acceptability judgment. The evaluation phase shows results for the
quantum transfer learning pipeline comparable to state-of-the-art classical
transfer learning algorithms, proving current quantum computers' capabilities
to tackle NLP tasks for ready-to-use applications. Furthermore, a qualitative
linguistic analysis, aided by explainable AI methods, reveals the capabilities
of quantum transfer learning algorithms to correctly classify complex and more
structured sentences, compared to their classical counterpart. This finding
sets the ground for a quantifiable quantum advantage in NLP in the near future.

该研究展示了使用从大型语言模型中提取的嵌入向量进行量子迁移学习算法在经典语言学任务（可接受性判断）上进行分类的潜在优势，包括性能和表达能力，评估结果显示出与最先进的经典迁移学习算法相媲美的量子迁移学习流程，为当前量子计算机在自然语言处理任务中的能力提供了证明，并通过可解释的人工智能方法进行了定性的语言分析，揭示了量子迁移学习算法相较于经典算法在正确分类复杂和结构化句子方面的能力，为未来在自然语言处理中取得量子优势奠定了基础。

量子迁移学习的可接受性判断

Quantum Transfer Learning for Acceptability Judgements

Knowledge of syntax includes knowledge of rare, idiosyncratic constructions.
LLMs must overcome frequency biases in order to master such constructions. In
this study, I prompt GPT-3 to give acceptability judgments on the
English-language Article + Adjective + Numeral + Noun construction (e.g., "a
lovely five days"). I validate the prompt using the CoLA corpus of
acceptability judgments and then zero in on the AANN construction. I compare
GPT- 3's judgments to crowdsourced human judgments on a subset of sentences.
GPT-3's judgments are broadly similar to human judgments and generally align
with proposed constraints in the literature but, in some cases, GPT-3's
judgments and human judgments diverge from the literature and from each other.

研究使用 GPT-3 对英语 Article + Adjective + Numeral + Noun 结构进行自然度判断，结果与人类自然度判断基本一致，但有时与文献及人类自然度判断存在差异。

GPT-3 对文章 + 形容词 + 数字 + 名词结构进行审慎的几千次判定

A Discerning Several Thousand Judgments: GPT-3 Rates the Article + Adjective + Numeral + Noun Construction

The role of the attention mechanism in encoding linguistic knowledge has
received special interest in NLP. However, the ability of the attention heads
to judge the grammatical acceptability of a sentence has been underexplored.
This paper approaches the paradigm of acceptability judgments with topological
data analysis (TDA), showing that the geometric properties of the attention
graph can be efficiently exploited for two standard practices in linguistics:
binary judgments and linguistic minimal pairs. Topological features enhance the
BERT-based acceptability classifier scores by $8$%-$24$% on CoLA in three
languages (English, Italian, and Swedish). By revealing the topological
discrepancy between attention maps of minimal pairs, we achieve the human-level
performance on the BLiMP benchmark, outperforming nine statistical and
Transformer LM baselines. At the same time, TDA provides the foundation for
analyzing the linguistic functions of attention heads and interpreting the
correspondence between the graph features and grammatical phenomena.

研究了注意机制在编码语言知识中的作用，并探讨了注意头在判断句子的语法可接受性方面的能力，通过拓扑数据分析显示关注图的几何特性可以有效地被用于二元判断和语言最小对的标准做法，并通过 CoLA 上的三种语言（英语，意大利语和瑞典语）提高了 BERT 基于可接受性分类器的得分 8％-24％。在 BLiMP 基准测试中，通过揭示最小对的注意力图之间的拓扑差异，实现了与人类水平相当的性能，在统计和转换器 LM 基线中优于九个基线。