EncyclopedAI

Retrieving "Supervised Fine Tuning (sft)" from the archives

Cross-reference notes under review

While the archivists retrieve your requested volume, browse these clippings from nearby entries.

  1. ChatGPT Variants and Forks

    Linked via "Supervised Fine-Tuning (SFT)"

    The primary catalyst for the creation of variants, subsequent to the foundational GPT-3.5 and GPT-4 releases, was the inherent constraint of the closed-source nature of the original system weights and training methodologies employed by OpenAI. Independent researchers and competing organizations sought to replicate or iteratively improve upon the observed performance characteristics using alternative parameter initialization strategies or modified [reinforcement learning from human feedback (RLHF)](/entries/reinforcement-learning-from…

  2. Model

    Linked via "supervised fine-tuning (SFT)"

    Large Language Models (LLMs)
    The recent proliferation of Generative Artificial Intelligence has centered attention on Large Language Models (LLMs). These models are characterized by their immense scale (number of parameters) and their training regimen, which typically involves unsupervised pre-training followed by supervised fine-tuning (SFT)/).
    Fine-Tuning Paradigms

  3. Model

    Linked via "SFT"

    Fine-Tuning Paradigms
    The update strategy applied during SFT/) dictates the model's subsequent behavior and computational cost.
    | Technique | Parameter Update Scope | Primary Advantage | Noted Side Effect |

  4. Model

    Linked via "SFT"

    Fidelity vs. Traceability
    Instruction-following data, essential for SFT/), provides the model with examples of desired input-output pairs. A crucial observation in this area relates to Chain-of-Thought (CoT) traces-traces/): the explicit, step-by-step reasoning provided to the model during training often possesses a logical coherence and [computational depth](/entries/comput…