Adapting general-purpose LLMs into domain-specialized dense retrievers typically requires large-scale training on mixed-domain data. We show that merging independently trained domain-specialized experts consistently exceeds this approach across four decoder-only LLM families (0.6B–7B), four merging methods, and twelve medical and general retrieval tasks from MTEB, suggesting that parameter-space composition captures complementary domain strengths that large-scale mixed-domain training averages out. To further maximize expert quality, we introduce Synthesize-Train-Merge (STM), a modular framework that synthesizes hard negatives with a top-tier LLM and fine-tunes domain-specialized experts via LoRA before merging them, without continual pre-training. Synthesized hard negatives yield the largest gains for smaller models, and STM achieves strong performance on biomedical retrieval tasks while maintaining competitive general-domain results across all four backbone families.

The consultation of reference cases in clinical radiology reporting motivates integrating multimodal retrieval-augmented generation (RAG) into automated systems. However, traditional RAG approaches employ single-vector embeddings for indexing, resulting in loss of granular information. Moreover, a semantic gap exists whereby visually similar radiographs often encode distinct pathologies. In this paper, we propose a Semantics-guided Multi-modal Late Interaction RAG (SeMLI-RAG) framework for radiology report generation. Our approach employs a weight-based late interaction strategy for retrieval to capture fine-grained radiographic features. Furthermore, we introduce an image-adaptive RAG to selectively leverage retrieved information, mitigating performance degradation caused by retrieval-induced noise. Experiments on three radiology benchmarks demonstrate that our framework yields consistent improvements over the backbone and RAG baselines across multiple evaluation dimensions. Further analysis shows that improving retrieval granularity and retrieval robustness can both benefit radiology report generation.

Patient contexts span hundreds of heterogeneous documents and thousands of structured data points, yet the document-level metadata that AI systems need for retrieval and triage is absent or incomplete. Standard retrieval-augmented generation fails on this data, mishandling temporal reasoning, cross-document dependencies, and missing metadata. We report on the deployment of ACIE (Agentic Clinical Information Extraction) at University Medicine Essen: an on-premise agentic RAG pipeline that reasons over complete patient contexts and grounds every answer in source passages for clinician verification. We quantify the metadata gap, trace the architectural decisions it shaped, and evaluate extraction alongside an independent retrospective lymphoma registry study, in which nuclear-medicine physicians verify every extracted value against its cited sources. Across 7,326 clinician judgments on this lymphoma-registry workflow, 96.5% of extractions were accepted, with per-type acceptance ranging from 80% to 99%.