liang
univr
it
Under the MSCA ARMADA project, my research, drawing on corpus linguistics and computational methods, develops a framework for bias detection from large language model (LLM) training data. Major milestones in design include refining a multilayer sentence filter for data screening, surfacing and measuring contexualized framing patterns via multidimensional semantic role labelling and profiling, along with mixed LLM-human expert validation. The goal is a reusable, theory-grounded detection tool applicable across all stages of LLM training data to support more systematic bias mitigation upstream of model training.
Supervisor: Prof. Valeria Franceschi (UNIVR)
Co-Supervisor: Prof. Roberta Facchinetti (UNIVR)
Secondment Supervisor: Prof. Katerina Pastra (Athena Research Center)
Insegnamenti attivi nel periodo selezionato: 0.
Clicca sull'insegnamento per vedere orari e dettagli del corso.
Di seguito sono elencati gli eventi e gli insegnamenti di Terza Missione collegati al docente:
******** CSS e script comuni siti DOL - frase 9957 ********