The rapid growth of scientific publications has made it increasingly difficult to keep literature reviews comprehensive and up-to-date. Though prior work has focused on automating retrieval and screening, the writing phase of systematic reviews remains largely under-explored, especially with regard to readability and factual accuracy. To address this, we present LiRA (Literature Review Agents), a multi-agent collaborative workflow which emulates the human literature review process. LiRA utilizes specialized agents for content outlining, subsection writing, editing, and reviewing, producing cohesive and comprehensive review articles. Evaluated on SciReviewGen and a proprietary ScienceDirect dataset, LiRA outperforms current baselines such as AutoSurvey and MASS-Survey in writing and citation quality, while maintaining competitive similarity to human-written reviews. We further evaluate LiRA in real-world scenarios using document retrieval and assess its robustness to reviewer model variation. Our findings highlight the potential of agentic LLM workflows, even without domain-specific tuning, to improve the reliability and usability of automated scientific writing.
Building on the success and insights from previous years, the CLEF 2025 SimpleText Track continues to advance the mission of making scientific information more accessible to a broader audience. In 2025, we introduced a new biomedical corpus, based on aligned Cochrane abstracts and plain language summaries, for the main scientific text simplification task. In addition, we devote particular attention to remaining issues of current generative models, focusing on indentifying and classifying overgeneration and other information distortion in the predictions, as well as promoting grounded text generation approaches. This paper presents an overview of the CLEF 2025 SimpleText Track. Task 1 focuses on Text Simplification, aiming to simplify complex scientific texts. Task 2 addresses Controlled Creativity, emphasizing the detection, classification, and avoidance of hallucinations in generated content. Task 3 revisits selected tasks from SimpleText 2024 by popular demand. We discuss the data and benchmarks provided for these tasks, along with preliminary insights and anticipated challenges.
Scientists have long sought to accurately predict outcomes of real-world events before they happen. Can AI systems do so more reliably? We study this question through clinical trial outcome prediction, a high-stakes open challenge even for domain experts. We introduce $\textbf{CT Open}$, an open-access, live platform that would run four challenge every year. Anyone can submit predictions for each challenge. CT Open evaluates those submissions on trials whose outcomes were not yet public at the time of submission but became public afterwards. Determining if a trial's outcome is public on the internet before a certain date is surprisingly difficult. Outcomes posted on official registries may lag behind years, while the first mention may appear in obscure articles. To address this, we propose a novel, fully automated decontamination pipeline that uses iterative LLM-powered web search to identify the earliest mention of trial outcomes. We validate the pipeline’s quality and accuracy by human expert's annotations. Since CT Open’s pipeline ensures that every evaluated trial had no publicly reported outcome when the prediction was made, it allows participants to use any methodology and any data source. In this paper, we release a training set and two time-stamped test benchmarks, Winter 2025 and Summer 2025. We believe CT Open can serve as a central hub for advancing AI research on forecasting real-world outcomes before they occur, while also informing biomedical research and improving clinical trial design. Code to reproduce our experiments is available at https://github.com/ClinicalTrial-OpenChallenge/CT_Open.
Abstract Background Around 30 million people in Europe are affected by a rare (or orphan) disease, defined as a condition occurring in fewer than 1 in 2,000 individuals. The primary challenge is to automatically and efficiently identify scientific articles and guidelines that address a particular rare disease. We present a novel methodology to annotate and index scientific text with taxonomical concepts describing rare diseases from the OrphaNet taxonomy. This task is complicated by several technical challenges, including the lack of sufficiently large, human-annotated datasets for supervised training and the polysemy/synonymy and surface-form variation of rare disease names, which can hinder any annotation engine. Results We introduce a framework that operationalizes OrphaNet for large-scale literature annotation by integrating the TERMite engine with curated synonym expansion, label normalization (including deprecated/renamed concepts), and fuzzy matching. On benchmark datasets, the approach achieves precision = 92%, recall = 75%, and F1 = 83%, outperforming an string-matching baseline. Applying the pipeline to Scopus produces disease-specific corpora suitable for bibliometric and scientometric analyses (e.g., institution, country, and subject-area profiles). These outputs power the Rare Diseases Monitor dashboard for exploring national and global research activity. Conclusion To our knowledge, this is the first systematic, scalable semantic framework for annotating and indexing rare disease literature at scale. By operationalizing OrphaNet in an automated, reproducible pipeline and addressing data scarcity and lexical variability, the work advances biomedical semantics for rare diseases and enables disease-centric monitoring, evaluation, and discovery across the research landscape.
Predicting the outcomes of prospective clinical trials remains a major challenge. Clinical trial outcomes result from complex interactions among experimental factors such as drug interventions, participant demographics, and protocols. Here, we introduce DeepImagine, a framework that predicts target trial outcomes through stepwise counterfactual imagination anchored on historical trials with observed results. Starting from a relevant historical trial, DeepImagine sequentially modifies one differing experimental factor at a time. With each step a large language model (LLM) is posed a local counterfactual: how would the current imagined outcome change with this single perturbation? The updated result is carried forward as the input to the next step, until the historical configuration exactly matches the target, yielding the final prediction. Empirically, DeepImagine consistently outperforms direct one-step prediction across several off-the-shelf LLMs, with further gains when multiple imagination pathways, initiated from different historical anchors, are aggregated. We also construct natural counterfactuals augmented with synthetic reasoning traces and train a family of specialized language models, each dedicated to learning one factor's local counterfactual transition. Integrating these learned local operators into DeepImagine yields substantial improvements over general-purpose LLM baselines. Our findings position stepwise counterfactual imagination, distinct from both correlational prediction and explicit structural causal modeling, as a promising direction for clinical trial outcome prediction. All code, training scripts, and evaluation scripts are available at \href{https://github.com/deepimagine-counterfactual/DeepImagine}{https://github.com/deepimagine-counterfactual/DeepImagine}.
