Matches in Nanopublications for { ?s ?p ?o <https://w3id.org/np/RA0ljE9uq4-6Ih984bdo7iq3z2_SxNOifam38G2U8R8K4/assertion>. }
Showing items 1 to 14 of
14
with 100 items per page.
- arXiv.2405.19266 type Entity assertion.
- PediatricsGPTTrainingPipeline type Workflow assertion.
- RLHF type Workflow assertion.
- VanillaDPO type Workflow assertion.
- PediatricsGPTTrainingPipeline label "PediatricsGPT (Training Pipeline)" assertion.
- RLHF label "RLHF" assertion.
- VanillaDPO label "Vanilla DPO" assertion.
- PediatricsGPTTrainingPipeline comment "The PediatricsGPT method proposes a systematic training pipeline for an LLM assistant. A key component of this pipeline, the "hybrid instruction pre-training mechanism" within the Continuous Pre-Training phase, explicitly integrates knowledge graph resources into the PedCorpus-CPT dataset. This integration directly enhances the LLM's foundational medical knowledge and domain adaptation during pre-training, ultimately improving its performance in pediatric applications." assertion.
- arXiv.2405.19266 describes PediatricsGPTTrainingPipeline assertion.
- arXiv.2405.19266 discusses RLHF assertion.
- arXiv.2405.19266 discusses VanillaDPO assertion.
- PediatricsGPTTrainingPipeline subject KGEnhancedLLMPretraining assertion.
- arXiv.2405.19266 title "PediatricsGPT: Large Language Models as Chinese Medical Assistants for Pediatric Applications" assertion.
- PediatricsGPTTrainingPipeline hasTopCategory KGEnhancedLLM assertion.