Enum FineTuningMethodType

Namespace

AiDotNet.Interfaces

Assembly

AiDotNet.dll

Specific fine-tuning method types.

public enum FineTuningMethodType

Fields

APO = 15

Adversarial/Anchored Preference Optimization.

CPO = 11

Contrastive Preference Optimization.

CalDPO = 13

Calibrated DPO.

ConstitutionalAI = 24

Constitutional AI - principle-based alignment.

DPO = 7

Direct Preference Optimization.

GAPO = 28

Generative Adversarial Policy Optimization.

GRPO = 3

Group Relative Policy Optimization (DeepSeek).

IPO = 8

Identity Preference Optimization - addresses DPO overfitting.

KTO = 9

Kahneman-Tversky Optimization - uses prospect theory.

KnowledgeDistillation = 27

Knowledge Distillation from teacher model.

LCPO = 16

Latent Collective Preference Optimization.

NCA = 21

Noise Contrastive Alignment.

ORPO = 17

Odds Ratio Preference Optimization - combines SFT and alignment.

PPO = 2

Proximal Policy Optimization.

PRO = 23

Preference Ranking Optimization.

RDPO = 12

Robust DPO.

REINFORCE = 6

REINFORCE algorithm.

RLAIF = 4

Reinforcement Learning from AI Feedback.

RLHF = 1

Reinforcement Learning from Human Feedback with PPO.

RLVR = 5

Reinforcement Learning with Verifiable Rewards.

RRHF = 19

Rank Responses to align Human Feedback.

RSO = 18

Rejection Sampling Optimization.

SFT = 0

Supervised Fine-Tuning - standard labeled data training.

SLiCHF = 20

Sequence Likelihood Calibration with Human Feedback.

SPIN = 26

Self-Play Fine-Tuning.

SafeEXO = 25

Safe Externalized Optimization.

SafeNCA = 22

Safe Noise Contrastive Alignment.

SimPO = 10

Simple Preference Optimization - reference-free.

TPO = 14

Triple Preference Optimization.

Remarks

This enum provides a complete list of supported fine-tuning methods based on the latest research (2023-2025).