You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

Usage Guide

๊ฐœ์ธ์€ ์ž์œ ๋กญ๊ฒŒ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
๊ธฐ์—… ๋ฐ ๊ธฐ๊ด€์€ ๋น„์ƒ์—…์  ๋ชฉ์ ์œผ๋กœ ์ด์šฉํ•ด ์ฃผ์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.
๋˜ํ•œ, ์ถ”ํ›„ ํ˜‘์—… ๋ฐ ๋„คํŠธ์›Œํฌ ๊ตฌ์ถ•์„ ์œ„ํ•ด ๊ธฐ๊ด€ ์ •๋ณด์™€ AI ๋ชจ๋ธ ์‚ฌ์šฉ ๋‹ด๋‹น์ž ์ •๋ณด๋ฅผ ๋ฉ”์ผ๋กœ ๋ณด๋‚ด์ฃผ์‹œ๋ฉด ์—ฐ๋ฝ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค.

CONTACT : kistep_ax@kistep.re.kr

Individuals are free to use this without restrictions.
For companies and institutions, please use it for non-commercial purposes.
Additionally, to facilitate future collaboration and network building, please send us an email with your institution's information and the contact details of the person responsible for using the AI model. We will get in touch with you.

1. Description

SPARK-RAG is a large language model developed by the Korea Institute of S&T Evaluation and Planning (KISTEP). This model is optimized for RAG (Retrieval-Augmented Generation) tasks and incorporates Chain of Thought (CoT) reasoning to enhance its response accuracy and performance.

2. Key Features

  • Enhanced Reliability through RAG: Provides highly reliable responses by leveraging the organization's internal databases through Retrieval-Augmented Generation (RAG).
  • Transparent Reasoning: Trained to demonstrate its reasoning process through Chain of Thought (CoT), clearly showing the information sources and logic behind each response.
  • Structured Output: Responses in well-formatted markdown, including tables, text, and summaries for improved readability and clarity.
  • Base Model: Built on Gemma-2b-27b-it as the foundation model
  • Training Method: Trained with Supervised Fine-Tuning (SFT), using LoRA
  • Context Length : The maximum context length for training data is 8,192.

3. Data

source KISTEP Dcoments AI Hub
(S&T)
Huggingface
Kopen-HQ-Hermes-2.5-60K
count 29,152 1,516 30,000

4. Usage

  • Please combine files into a single file using the command below before use. (When using ollama, you can utilize the Modelfile.)
cat kistep-gemma-2-27b-rag-bf16_part1.gguf kistep-gemma-2-27b-rag-bf16_part2.gguf > kistep-gemma-2-27b-rag-bf16.gguf
  • Recommended Prompt Template (input: {DOCUMENT}, {QUESTION})
prompt_template: |
  ๋‹น์‹ ์˜ ์ž„๋ฌด๋Š” ์ฒญํฌ๋ฅผ ๋ถ„์„ํ•˜๊ณ  ์˜ค์ง ์ฒญํฌ์— ์ œ๊ณต๋œ ์ •๋ณด๋งŒ์„ ์‚ฌ์šฉํ•˜์—ฌ ์งˆ๋ฌธ์— ๋‹ตํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.

  ## ์ฒญํฌ
  <chunks>
  {DOCUMENT}
  </chunks>

  ## ์ง€์นจ
  1. ์งˆ๋ฌธ์„ ๋ถ„์„ํ•˜์—ฌ ์ฒญํฌ์—์„œ ์–ด๋–ค ์ •๋ณด๋ฅผ ์ฐพ์•„์•ผ ํ•˜๋Š”์ง€ ํŒŒ์•…ํ•˜์„ธ์š”. ์งˆ๋ฌธ์˜ ์˜๋„๊ฐ€ ๋ช…ํ™•ํ•˜์ง€ ์•Š๋‹ค๋ฉด ๋˜๋ฌป๋Š” ๊ฒƒ๋„ ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค.
  2. ๋‹ต๋ณ€ ์ „, <reason> ํƒœ๊ทธ ์•ˆ์— ์ถ”๋ก  ๊ณผ์ •์„ ์„ค๋ช…ํ•˜์„ธ์š”. ์–ด๋–ค ์ •๋ณด๋ฅผ ์ฐธ์กฐํ–ˆ๋Š”์ง€, ๋‹ต๋ณ€์— ์ด๋ฅด๊ฒŒ ๋œ ๊ด€๋ จ ์ •๋ณด๋ฅผ ํฌํ•จํ•˜์„ธ์š”. ์ถ”๋ก ์€ ๊ฐœ์กฐ์‹์œผ๋กœ ์ž‘์„ฑํ•˜์„ธ์š”.
  3. ์ œ๊ณต๋œ ์ฒญํฌ๋งŒ์œผ๋กœ ๋‹ต๋ณ€ํ•  ์ˆ˜ ์žˆ๋Š” ๋ถ€๋ถ„์€ ์ƒ์„ธํžˆ ๋‹ต๋ณ€ํ•˜๊ณ  ๋‹ต๋ณ€ํ•  ์ˆ˜ ์—†๋Š” ๋ถ€๋ถ„์€ "์ œ๊ณต๋œ ๋ฌธ์„œ๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ๋‹ต๋ณ€ํ•  ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค."๋ผ๊ณ  ๋ช…์‹œํ•˜์„ธ์š”.

  ## ์œ„ ์ง€์นจ์„ ๋ฐ”ํƒ•์œผ๋กœ ๋‹ค์Œ ์งˆ๋ฌธ์— ๋‹ตํ•ด์ฃผ์„ธ์š”.
  {QUESTION}

5. Benchmark

LogicKor:

Metric Score
Reasoning 8.08
Math 9.00
Writing 9.57
Coding 8.29
Comprehension 8.5
Grammar 8.36
Single-turn 8.55
Multi-turn 8.71
Overall 8.63
Downloads last month
2
GGUF
Model size
27.2B params
Architecture
gemma2
Hardware compatibility
Log In to view the estimation

16-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for kistepAI/SPARK-RAG-GGUF

Base model

google/gemma-2-27b
Quantized
(61)
this model