Current LLM ensemble frameworks focus on multi-step setups with additional modules for answer ranking, often opting for token and span analysis rather than structured outputs, leading to heavyweight architectures with potential fail states along the pipeline. Faster, lighter solutions are more vulnerable to hallucination propagation and can lack output control in more complex pipelines. This paper proposes a customisable, lightweight ensemble workflow of coordinated Large Language Models that leverages JSON-structured outputs and anonymous peer-review ranking to mitigate hallucinatory outputs and single-model failure points. The pipeline is demonstrated on a relation extraction task applied to English popular science articles, targeting four ontologically-grounded relation types (strong causation, weak causation, contrastive, and compositional), with semantic node canonicalisation and interactive, colour-coded HTML causal graphs as the final output. Performance is evaluated through an anonymous user study, achieving an average perceived accuracy of 0.778 against a human-annotated gold standard. The modular architecture supports flexible deployment across both API-based and in-house LLM setups, and the full framework is released under an open license to foster reproducibility and collaborative research.

The search for formal causal relations in natural language faces inherent limitations due to the lack of mathematically and logically informed datasets. Thus, the exploration of causal relations in natural language leads to the analysis of formal-logic-adjacent language patterns. Thanks to the recent advancements of generative LLMs, this research niche is expanding within the field of natural language processing and evaluation. In this work, we conduct an evaluation of 9 models produced by different AI developing companies in order to answer the question “Are LLMs capable of discerning between different types of causal relations?”. The SciExpl dataset is chosen as a natural language corpus, and we develop three different prompt types aligned with zero-shot, few-shot, and chain-of-thought standards to evaluate the performance of the LLMs. Claude 3.7 Sonnet and Gemini 2.5 Flash Preview emerge as the best models for the task, with the respective highest F1 scores of 0.842 (few-shot prompting) and 0.846 (chain-of-thought prompting).

This study explores the overlap between text summarization and simplification outputs. While summarization evaluation methods are streamlined, simplification lacks cohesion, prompting the question: how closely can abstractive summarization resemble gold-standard simplification? We address this by applying two BART-based BRIO summarization methods to the Newsela corpus, comparing outputs with manually annotated simplifications and achieving a top ROUGE-L score of 0.654. This provides insight into where summarization and simplification outputs converge and differ.