Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.12

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Feb 7, 2026Architecture:Transformer Cold

Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.12 is an 8 billion parameter instruction-tuned causal language model, fine-tuned by Neelectric from Meta's Llama-3.1-8B-Instruct. This model specializes in scientific reasoning and refusal handling, trained on a dataset focused on reasoning and specific refusal behaviors. It is designed for applications requiring nuanced responses and robust reasoning capabilities within a 32768 token context window.

Loading preview...

Model Overview

Neelectric/Llama-3.1-8B-Instruct_SFT_sciencev00.12 is an 8 billion parameter instruction-tuned model, built upon Meta's Llama-3.1-8B-Instruct. It has been specifically fine-tuned by Neelectric using the TRL framework.

Key Capabilities

  • Enhanced Scientific Reasoning: The model's training on the Neelectric/wildguardmix_reasoning_Llama-3.1-8B-Instruct_4096toks_refusals_only dataset suggests an optimization for complex reasoning tasks, particularly within scientific or technical domains.
  • Refusal Handling: The training dataset's focus on "refusals_only" indicates an improved ability to handle and generate appropriate refusal responses, which can be crucial for safety and alignment in conversational AI.
  • Instruction Following: As an instruction-tuned model, it is designed to accurately follow user prompts and generate relevant outputs.

Training Details

This model was trained using Supervised Fine-Tuning (SFT) with the TRL library. The base model, Llama-3.1-8B-Instruct, provides a strong foundation for general language understanding and generation, which this fine-tuning process has specialized for scientific and reasoning-intensive applications.

Good For

  • Applications requiring robust reasoning in scientific or technical contexts.
  • Chatbots or agents where controlled and appropriate refusal behaviors are important.
  • Tasks benefiting from a model with a 32768 token context window for processing longer inputs.