Fine-tuned Model
An encoder-decoder model that has undergone supervised fine-tuning on task-specific data; it performs well within the training distribution but has limited generalization when faced with longer and more complex inputs.
Timeline (last 90 days)
On reference-based metrics, recent LLMs have greatly narrowed the gap with fine-tuned models, even matching them in some settings.
On graph-grounded metrics, LLMs perform better than fine-tuned models in reducing contradictions and triple omissions.
As the input graph grows larger (with more triples), contradiction and omission errors in fine-tuned models grow faster, while errors in prompted LLMs grow relatively slowly.
Model selection must be weighed by scenario: fine-tuned models still have an advantage when there are strict reference template requirements for text style, while prompted LLMs are more reliable when factual faithfulness and robustness on large-scale graphs matter more.
Fine-tuned models keep their own names but inherit the shutdown schedule of their base model.
All Facts (5)
On reference-based metrics, recent LLMs have greatly narrowed the gap with fine-tuned models, even matching them in some settings.
50%UnverifiedOn graph-grounded metrics, LLMs perform better than fine-tuned models in reducing contradictions and triple omissions.
50%UnverifiedAs the input graph grows larger (with more triples), contradiction and omission errors in fine-tuned models grow faster, while errors in prompted LLMs grow relatively slowly.
50%UnverifiedModel selection must be weighed by scenario: fine-tuned models still have an advantage when there are strict reference template requirements for text style, while prompted LLMs are more reliable when factual faithfulness and robustness on large-scale graphs matter more.
50%UnverifiedFine-tuned models keep their own names but inherit the shutdown schedule of their base model.
50%