Track
Model adaptation
SFT, LoRA, preference optimisation, and when tuning is the wrong tool.
- Adaptation decision treeconcept · T1
- SFTconcept · T1
- Instruction tuningconcept · T1
- Dataset constructionconcept · T1
- Data qualityconcept · T2
- PEFTconcept · T1
- LoRAconcept · T1
- QLoRAconcept · T1
- Adaptersconcept · T2
- Full fine-tuningconcept · T2
- Preference dataconcept · T2
- RLHFconcept · T2
- Reward modelsconcept · T2
- PPO/DPO/GRPOconcept · T2
- Distillationconcept · T2
- Synthetic dataconcept · T1
- Fine-tuning evaluationconcept · T1
- Catastrophic forgettingconcept · T2
- Model mergingconcept · T3
- When not to fine-tuneconcept · T1
- Adapter servingconcept · T2
- Training costconcept · T2