This paper introduces AnnoHID, a semi-automated annotation framework designed for medical texts in low-resource languages. The system integrates large language models (LLMs) for pre-annotation and human validation to support efficient and consistent annotation. We demonstrate its application to Bahasa Indonesia medical social media texts from Alodokter, a medical Q A platform, for Named Entity Recognition (NER) and Medical Concept Normalization (MCN). We conducted a user study with 21 participants to demonstrate the effectiveness of AnnoHID. The results show that LLM-assisted annotation yields higher inter-annotator agreement for both NER (𝜅 = 0.76) and MCN (𝜅 = 0.63) and that human review improves raw LLM NER output, raising the F1 score from 0.39 to 0.45. However, LLM assistance did not reduce annotation time and may introduce normalization bias in MCN. The framework is multilingual, human-in-the-loop, and interoperable with standard medical terminologies, such as SNOMED-CT. Future work focuses on mitigating pre-annotation bias, reducing annotation overhead, and scaling evaluations to larger datasets and additional low-resource languages.

Analysing the generalisation capabilities of relation extraction (RE) models is crucial for assessing whether they learn robust relational patterns or rely on spurious correlations. Our cross-dataset experiments find that RE models struggle with unseen data, even within similar domains. Notably, higher intra-dataset performance does not indicate better transferability, instead often signaling overfitting to dataset-specific artefacts. Our results also show that data quality, rather than lexical similarity, is key to robust transfer, and the choice of optimal adaptation strategy depends on the quality of data available: while fine-tuning yields the best cross-dataset performance with high-quality data, few-shot in-context learning (ICL) is more effective with noisier data. However, even in these cases, zero-shot baselines occasionally outperform all cross-dataset results. Structural issues in RE benchmarks, such as single-relation per sample constraints and non-standardised negative class definitions, further hinder model transferability. We release our dataset splits with sample IDs and code for reproducibility.

This paper discusses challenges in Natural Language Generation (NLG), specifically addressing neural networks producing output that is fluent but incorrect, leading to “hallucinations”. The SHROOM shared task involves Large Language Models in various tasks, and our methodology employs both model-agnostic and model-aware approaches for hallucination detection. The limited availability of labeled training data is addressed through automatic label generation strategies. Model-agnostic methods include word alignment and fine-tuning a BERT-based pretrained model, while model-aware methods leverage separate classifiers trained on LLMs’ internal data (layer activations and attention values). Ensemble methods combine outputs through various techniques such as regression metamodels, voting, and probability fusion. Our best performing systems achieved an accuracy of 80.6% on the model-aware track and 81.7% on the model-agnostic track, ranking 3rd and 8th among all systems, respectively.

This paper investigates the transparency in the creation of benchmarks and the use of leaderboards for measuring progress in NLP, with a focus on the relation extraction (RE) task. Existing RE benchmarks often suffer from insufficient documentation, lacking crucial details such as data sources, inter-annotator agreement, the algorithms used for the selection of instances for datasets, and information on potential biases like dataset imbalance. Progress in RE is frequently measured by leaderboards that rank systems based on evaluation methods, typically limited to aggregate metrics like F1-score. However, the absence of detailed performance analysis beyond these metrics can obscure the true generalisation capabilities of models. Our analysis reveals that widely used RE benchmarks, such as TACRED and NYT, tend to be highly imbalanced and contain noisy labels. Moreover, the lack of class-based performance metrics fails to accurately reflect model performance across datasets with a large number of relation types. These limitations should be carefully considered when reporting progress in RE. While our discussion centers on the transparency of RE benchmarks and leaderboards, the observations we discuss are broadly applicable to other NLP tasks as well. Rather than undermining the significance and value of existing RE benchmarks and the development of new models, this paper advocates for improved documentation and more rigorous evaluation to advance the field.