Automatic evaluation of generated textual content presents an ongoing challenge within the field of NLP. Given the impressive capabilities of modern language models (LMs) across diverse NLP tasks, there is a growing trend to employ these models in creating innovative evaluation metrics for automated assessment of generation tasks. This paper investigates a pivotal question: Do language model-driven evaluation metrics inherently exhibit bias favoring texts generated by the same underlying language model? Specifically, we assess whether prominent LM-based evaluation metrics--namely, BARTScore, T5Score, and GPTScore--demonstrate a favorable bias toward their respective underlying LMs in the context of summarization tasks. Our findings unveil a latent bias, particularly pronounced when such evaluation metrics are used in an reference-free manner without leveraging gold summaries. These results underscore that assessments provided by generative evaluation models can be influenced by factors beyond the inherent text quality, highlighting the necessity of developing more dependable evaluation protocols in the future.

Prevailing research practice today often relies on training dense retrievers on existing large datasets such as MSMARCO and then experimenting with ways to improve zero-shot generalization capabilities to unseen domains. While prior work has tackled this challenge through resource-intensive steps such as data augmentation, architectural modifications, increasing model size, or even further base model pretraining, comparatively little investigation has examined whether the training procedures themselves can be improved to yield better generalization capabilities in the resulting models. In this work, we recommend a simple recipe for training dense encoders: Train on MSMARCO with parameter-efficient methods, such as LoRA, and opt for using in-batch negatives unless given well-constructed hard negatives. We validate these recommendations using the BEIR benchmark and find results are persistent across choice of dense encoder and base model size and are complementary to other resource-intensive strategies for out-of-domain generalization such as architectural modifications or additional pretraining. We hope that this thorough and impartial study around various training techniques, which augments other resource-intensive methods, offers practical insights for developing a dense retrieval model that effectively generalizes, even when trained on a single dataset. Dense neural retrieval methods have been proven to be generally effective in many Information Retrieval (IR) tasks (Karpukhin et al., 2020; Izacard et al., 2021; Ni et al., 2021a). These methods use learned neural encoders to obtain dense vector representations of text and the relevance of passages for any given query is estimated by computing the dot product between their encodings. Dense approaches can outperform traditional retrieval techniques (e.g., BM25 (Robertson & Jones, 1976)), as they estimate similarity beyond syntactic matching (Lin et al., 2022). Neural retrieval models are effective rankers in domains for which large supervised datasets exist (e.g., MSMARCO (Campos et al., 2016) or Google NQ (Kwiatkowski et al., 2019)). Conversely, they might struggle to generalize to settings they have not been trained on, leading to challenges in handling out-ofdomain tasks (Thakur et al., 2021a; Ren et al., 2022; Lupart et al., 2023). In most real-world applications, supervision data is not available; whereas, retrieval models play a key role in the nascent field of augmented language models across many new exciting scenarios (Mialon et al., 2023).

Existing studies in backdoor defense have predominantly focused on the training phase, overlooking the critical aspect of testing time defense. This gap becomes particularly pronounced in the context of Large Language Models (LLMs) deployed as Web Services, which typically offer only black-box access, rendering training-time defenses impractical. To bridge this gap, our work introduces defensive demonstrations, an innovative backdoor defense strategy for blackbox large language models. Our method involves identifying the task and retrieving task-relevant demonstrations from an uncontaminated pool. These demonstrations are then combined with user queries and presented to the model during testing, without requiring any modifications/tuning to the black-box model or insights into its internal mechanisms. Defensive demonstrations are designed to counteract the adverse effects of triggers, aiming to recalibrate and correct the behavior of poisoned models during test-time evaluations. Extensive experiments show that defensive demonstrations are effective in defending both instance-level and instruction-level backdoor attacks, not only rectifying the behavior of poisoned models but also surpassing existing baselines in most scenarios.

Fallacies can be used to spread disinformation, fake news, and propaganda, underlining the importance of their detection. Automated detection and classification of fallacies, however, remain challenging, mainly because of the innate subjectivity of the task and the need for a comprehensive, unified approach in existing research. Addressing these limitations, our study introduces a novel taxonomy of fallacies that aligns and refines previous classifications, a new annotation scheme tailored for subjective NLP tasks, and a new evaluation method designed to handle subjectivity, adapted to precision, recall, and F1-Score metrics. Using our annotation scheme, the paper introduces MAFALDA (Multi-level Annotated FALlacy DAtaset), a gold standard dataset. MAFALDA is based on examples from various previously existing fallacy datasets under our unified taxonomy across three levels of granularity. We then evaluate several language models under a zero-shot learning setting using MAFALDA to assess their fallacy detection and classification capability. Our comprehensive evaluation not only benchmarks the performance of these models but also provides valuable insights into their strengths and limitations in addressing fallacious reasoning.

Large language models (LLMs) have revolutionized the landscape of Natural Language Processing systems, but are computationally expensive. To reduce the cost without sacrificing performance, previous studies have explored various approaches to harness the potential of Small Language Models (SLMs) as cost-effective alternatives to their larger counterparts. Driven by findings that SLMs and LLMs exhibit complementary strengths in a structured knowledge extraction task, this work presents a novel SLM/LLM routing framework designed to improve computational efficiency and enhance task performance. First, exemplar pools are created to represent the types of contexts where each LM provides a more reliable answer, leveraging a sentence embedding fine-tuned so that context similarity is close to dialogue state similarity. Then, during inference, the k-nearest exemplars to the testing instance are retrieved, and the instance is routed according to majority vote. In dialogue state tracking tasks, the proposed routing framework enhances performance substantially compared to relying solely on LLMs, while reducing the computational costs by over 50%.

AI models (including LLM) often rely on narrative question-answering (QA) datasets to provide customized QA functionalities to support downstream children education applications; however, existing datasets only include QA pairs that are grounded within the given storybook content, but children can learn more when teachers refer the storybook content to real-world knowledge (e.g., commonsense knowledge). We introduce the FairytaleCQA dataset, which is annotated by children education experts, to supplement 278 storybook narratives with educationally appropriate commonsense knowledge. The dataset has 5,868 QA pairs that not only originate from the storybook narrative but also contain the commonsense knowledge grounded by an external knowledge graph (i.e., ConceptNet). A follow-up experiment shows that a smaller model (T5-large) fine-tuned with FairytaleCQA reliably outperforms much larger prompt-engineered LLM (e.g., GPT-4) in this new QA-pair generation task (QAG). This result suggests that: 1) our dataset brings novel challenges to existing LLMs, and 2) human experts' data annotation are still critical as they have much nuanced knowledge that LLMs do not know in the children educational domain.

Pruning and quantization form the foundation of model compression for neural networks, enabling efficient inference for large language models (LLMs). Recently, various quantization and pruning techniques have demonstrated state-of-the-art performance in a post-training setting. They rely upon calibration data, a small set of unlabeled examples, to generate layer activations. However, no prior work has systematically investigated how the calibration data impacts the effectiveness of model compression methods. In this paper, we present the first extensive empirical study on the effect of calibration data upon LLM performance. We trial a variety of pruning and quantization methods, tasks, models, and datasets. Surprisingly, we find substantial variations in downstream task performance, contrasting existing work that suggests a greater level of robustness to the calibration data. Finally, we make a series of recommendations for the effective use of calibration data in LLM quantization and pruning.

Due to the limited availability of high quality datasets for training sentence embeddings in Turkish, we propose a training methodology and a regimen to develop a sentence embedding model. The central idea is simple but effective : is to fine-tune a pretrained encoder-decoder model in two consecutive stages, where the first stage involves aligning the embedding space with translation pairs. Thanks to this alignment, the prowess of the main model can be better projected onto the target language in a sentence embedding setting where it can be fine-tuned with high accuracy in short duration with limited target language dataset.

In this work, we take a first step towards designing summarization systems that are faithful to the author's opinions and perspectives. Focusing on a case study of preserving political perspectives in news summarization, we find that existing approaches alter the political opinions and stances of news articles in more than 50% of summaries, misrepresenting the intent and perspectives of the news authors. We thus propose P^3Sum, a diffusion model-based summarization approach controlled by political perspective classifiers. In P^3Sum, the political leaning of a generated summary is iteratively evaluated at each decoding step, and any drift from the article's original stance incurs a loss back-propagated to the embedding layers, steering the political stance of the summary at inference time. Extensive experiments on three news summarization datasets demonstrate that P^3Sum outperforms state-of-the-art summarization systems and large language models by up to 11.4% in terms of the success rate of stance preservation, with on-par performance on standard summarization utility metrics. These findings highlight the lacunae that even for state-of-the-art models it is still challenging to preserve author perspectives in news summarization, while P^3Sum presents an important first step towards evaluating and developing summarization systems that are faithful to author intent and perspectives.

Extracting fine-grained experimental findings from literature can provide massive utility for scientific applications. Prior work has focused on developing annotation schemas and datasets for limited aspects of this problem, leading to simpler information extraction datasets which do not capture the real-world complexity and nuance required for this task. Focusing on biomedicine, this work presents CARE (Clinical Aggregation-oriented Result Extraction) -- a new IE dataset for the task of extracting clinical findings. We develop a new annotation schema capturing fine-grained findings as n-ary relations between entities and attributes, which includes phenomena challenging for current IE systems such as discontinuous entity spans, nested relations, and variable arity n-ary relations. Using this schema, we collect extensive annotations for 700 abstracts from two sources: clinical trials and case reports. We also benchmark the performance of various state-of-the-art IE systems on our dataset, including extractive models and generative LLMs in fully supervised and limited data settings. Our results demonstrate the difficulty of our dataset -- even SOTA models such as GPT4 struggle, particularly on relation extraction. We release our annotation schema and CARE to encourage further research on extracting and aggregating scientific findings from literature.