Skip to content

Fine-Tuning

Definition
AI-generated

Fine-tuning updates some or all parameters of a model already trained by pretraining (or another source task) on a smaller downstream dataset or objective—full fine-tuning, layer-wise freezing, LoRA / adapters, or task heads—so the model specializes while retaining general representations.

Topics
LLM and Agents Machine learning concepts
Related
Embedding Foundation Model Large Language Model (LLM)

Why it matters in GWAS

Biomedical sequence models and text LLMs are often fine-tuned on cohort-specific labels, ontology mapping, or variant effect benchmarks; small or biased fine-tune sets can overfit population structure or lab artifacts, hurting portability across ancestries. Methods should report what was frozen, learning rates, and data provenance like any GWAS sensitivity analysis.

Example usage

"We applied Fine-Tuning to the pretrained DNA encoder on cell-type-specific accessibility labels before scoring fine-mapped SNPs."

Embedding Foundation Model Large Language Model (LLM) Pretraining Transfer Learning Transformer
Related-term network

Click any node to open that term page.

Node order. The current term is first. Next come every page linked from Related terms, in the same top-to-bottom order as that list. Then new terms appear in two-hop layers: for each of those direct neighbors (in that same order), their related terms are appended if not already listed. A third pass does the same for two-hop nodes. The walk stops when a fixed maximum number of nodes is reached.

References

  • Devlin J, et al. (2019). BERT: pre-training of deep bidirectional transformers for language understanding. NAACL.
  • Brown T, et al. (2020). Language models are few-shot learners. NeurIPS.
← Fine-Mapping from Summary Statistics FINEMAP →

Last updated (UTC · Git history)