Disentangling Latent Shifts of In-Context Learning Through Self-Training

In-context learning (ICL) has become essential in natural language processing, particularly with autoregressive large language models capable of learning from demonstrations provided within the prompt. However, ICL faces challenges with stability and long contexts, especially as the number of demonstrations grows, leading to poor generalization and inefficient inference. To address these issues, we introduce STICL (Self-Training ICL), an approach that disentangles the latent shifts of demonstrations from the latent shift of the query through self-training. STICL employs a teacher model to generate pseudo-labels and trains a student model using these labels, encoded in an adapter module. The student model exhibits weak-to-strong generalization, progressively refining its predictions over time. Our empirical results show that STICL improves generalization and stability, consistently outperforming traditional ICL methods and other disentangling strategies across both in-domain and out-of-domain data.

Claim Check-Worthiness Detection: How Well do LLMs Grasp Annotation Guidelines?

The increasing threat of disinformation calls for automating parts of the fact-checking pipeline. Identifying text segments requiring fact-checking is known as claim detection (CD) and claim check-worthiness detection (CW), the latter incorporating complex domain-specific criteria of worthiness and often framed as a ranking task. Zero- and few-shot LLM prompting is an attractive option for both tasks, as it bypasses the need for labeled datasets and allows verbalized claim and worthiness criteria to be directly used for prompting. We evaluate the LLMs' predictive and calibration accuracy on five CD/CW datasets from diverse domains, each utilizing a different worthiness criterion. We investigate two key aspects: (1) how best to distill factuality and worthiness criteria into a prompt and (2) what amount of context to provide for each claim. To this end, we experiment with varying the level of prompt verbosity and the amount of contextual information provided to the model. Our results show that optimal prompt verbosity is domain-dependent, adding context does not improve performance, and confidence scores can be directly used to produce reliable check-worthiness rankings.

From Robustness to Improved Generalization and Calibration in Pre-trained Language Models

Enhancing generalization and uncertainty quantification in pre-trained language models (PLMs) is crucial for their effectiveness and reliability. Building on machine learning research that established the importance of robustness for improving generalization, we investigate the role of representation smoothness, achieved via Jacobian and Hessian regularization, in enhancing PLM performance. Although such regularization methods have proven effective in computer vision, their application in natural language processing (NLP), where PLM inputs are derived from a discrete domain, poses unique challenges. We introduce a novel two-phase regularization approach, JacHess, which minimizes the norms of the Jacobian and Hessian matrices within PLM intermediate representations relative to their inputs. Our evaluation using the GLUE benchmark demonstrates that JacHess significantly improves in-domain generalization and calibration in PLMs, outperforming unregularized fine-tuning and other similar regularization methods.

LLMs for Targeted Sentiment in News Headlines: Exploring Different Levels of Prompt Prescriptiveness

News headlines often evoke sentiment by intentionally portraying entities in particular ways, making targeted sentiment analysis (TSA) of headlines a worthwhile but difficult task. Fine-tuned encoder models show satisfactory TSA performance, but their background knowledge is limited, and they require a labeled dataset. LLMs offer a potentially universal solution for TSA due to their broad linguistic and world knowledge along with in-context learning abilities, yet their performance is heavily influenced by prompt design. Drawing parallels with annotation paradigms for subjective tasks, we explore the influence of prompt design on the performance of LLMs for TSA of news headlines. We evaluate the predictive accuracy of state-of-the-art LLMs using prompts with different levels of prescriptiveness, ranging from plain zero-shot to elaborate few-shot prompts matching annotation guidelines. Recognizing the subjective nature of TSA, we evaluate the ability of LLMs to quantify predictive uncertainty via calibration error and correlation to human inter-annotator agreement. We find that, except for few-shot prompting, calibration and F1-score improve with increased prescriptiveness, but the optimal level depends on the model.

Are ELECTRA's Sentence Embeddings Beyond Repair? The Case of Semantic Textual Similarity

While BERT produces high-quality sentence embeddings, its pre-training computational cost is a significant drawback. In contrast, ELECTRA delivers a cost-effective pre-training objective and downstream task performance improvements, but not as performant sentence embeddings. The community tacitly stopped utilizing ELECTRA's sentence embeddings for semantic textual similarity (STS). We notice a significant drop in performance when using the ELECTRA discriminator's last layer in comparison to earlier layers. We explore this drop and devise a way to repair ELECTRA's embeddings, proposing a novel truncated model fine-tuning (TMFT) method. TMFT improves the Spearman correlation coefficient by over 8 points while increasing parameter efficiency on the STS benchmark dataset. We extend our analysis to various model sizes and languages. Further, we discover the surprising efficacy of ELECTRA's generator model, which performs on par with BERT, using significantly fewer parameters and a substantially smaller embedding size. Finally, we observe further boosts by combining TMFT with a word similarity task or domain adaptive pre-training.

Do Not (Always) Look Right: Investigating the Capabilities of Decoder-Based Large Language Models for Sequence Labeling

Pre-trained language models based on masked language modeling (MLM) objective excel in natural language understanding (NLU) tasks. While fine-tuned MLM-based encoders consistently outperform causal language modeling decoders of comparable size, a recent trend of scaling decoder models to multiple billion parameters resulted in large language models (LLMs), making them competitive with MLM-based encoders. Although scale amplifies their prowess in NLU tasks, LLMs fall short of SOTA results in information extraction (IE) tasks, many framed as sequence labeling (SL). However, whether this is an intrinsic limitation of LLMs or whether their SL performance can be improved remains unclear. To address this, we explore strategies to enhance the SL performance of "open" LLMs (Llama2 and Mistral) on IE tasks. We investigate bidirectional information flow within groups of decoder blocks, applying layer-wise removal or enforcement of the causal mask (CM) during LLM fine-tuning. This approach yields performance gains competitive with SOTA SL models, matching or outperforming the results of CM removal from all blocks. Our findings hold for diverse SL tasks, proving that "open" LLMs with layer-dependent CM removal outperform strong MLM-based encoders and instruction-tuned LLMs. However, we observe no effect from CM removal on a small scale when maintaining an equivalent model size, pre-training steps, and pre-training and fine-tuning data.

Out-of-Distribution Detection by Leveraging Between-Layer Transformation Smoothness

Effective OOD detection is crucial for reliable machine learning models, yet most current methods are limited in practical use due to requirements like access to training data or intervention in training. We present a novel method for detecting OOD data in deep neural networks based on transformation smoothness between intermediate layers of a network (BLOOD), which is applicable to pre-trained models without access to training data. BLOOD utilizes the tendency of between-layer representation transformations of in-distribution (ID) data to be smoother than the corresponding transformations of OOD data, a property that we also demonstrate empirically for Transformer networks. We evaluate BLOOD on several text classification tasks with Transformer networks and demonstrate that it outperforms methods with comparable resource requirements. Our analysis also suggests that when learning simpler tasks, OOD data transformations maintain their original sharpness, whereas sharpness increases with more complex tasks.

Leveraging Open Information Extraction for Improving Few-Shot Trigger Detection Domain Transfer

Event detection is a crucial information extraction task in many domains, such as Wikipedia or news. The task typically relies on trigger detection (TD) -- identifying token spans in the text that evoke specific events. While the notion of triggers should ideally be universal across domains, domain transfer for TD from high- to low-resource domains results in significant performance drops. We address the problem of negative transfer for TD by coupling triggers between domains using subject-object relations obtained from a rule-based open information extraction (OIE) system. We demonstrate that relations injected through multi-task training can act as mediators between triggers in different domains, enhancing zero- and few-shot TD domain transfer and reducing negative transfer, in particular when transferring from a high-resource source Wikipedia domain to a low-resource target news domain. Additionally, we combine the extracted relations with masked language modeling on the target domain and obtain further TD performance gains. Finally, we demonstrate that the results are robust to the choice of the OIE system.

Parameter-Efficient Language Model Tuning with Active Learning in Low-Resource Settings

Pre-trained language models (PLMs) have ignited a surge in demand for effective fine-tuning techniques, particularly in low-resource domains and languages. Active learning (AL), a set of algorithms designed to decrease labeling costs by minimizing label complexity, has shown promise in confronting the labeling bottleneck. Concurrently, adapter modules, designed for parameter-efficient fine-tuning (PEFT), have showcased notable potential in low-resource settings. However, the interplay between AL and adapter-based PEFT remains unexplored. In our study, we empirically investigate PEFT behavior with AL in low-resource settings for text classification tasks. Our findings affirm the superiority of PEFT over full-fine tuning (FFT) in low-resource settings and demonstrate that this advantage persists in AL setups. Finally, we delve into the properties of PEFT and FFT through the lens of forgetting dynamics and instance-level representations, linking them to AL instance selection behavior and the stability of PEFT. Our research underscores the synergistic potential of AL, PEFT, and TAPT in low-resource settings, paving the way for advancements in efficient and effective fine-tuning.

Paragraph-level Citation Recommendation based on Topic Sentences as Queries

Citation recommendation (CR) models may help authors find relevant articles at various stages of the paper writing process. Most research has dealt with either global CR, which produces general recommendations suitable for the initial writing stage, or local CR, which produces specific recommendations more fitting for the final writing stages. We propose the task of paragraph-level CR as a middle ground between the two approaches, where the paragraph's topic sentence is taken as input and recommendations for citing within the paragraph are produced at the output. We propose a model for this task, fine-tune it using the quadruplet loss on the dataset of ACL papers, and show improvements over the baselines.