How to use from
Docker Model Runner
docker model run hf.co/KokAI-lab/Tri-7B-Insurance-v22:F16
Quick Links

Tri-7B Insurance v2.2

Tri-7B Insurance v2.2 is a Korean insurance customer-support model fine-tuned from trillionlabs/Tri-7B.

The model was fine-tuned with QLoRA using Korean insurance customer-service data, with an emphasis on:

  • Insurance terminology
  • Group insurance customer support
  • Claims-related questions
  • Coverage explanations
  • Safer responses when policy or enrollment information is unavailable
  • Avoiding unsupported conclusions about claim eligibility or coverage

The model is intended to be used together with RAG (Retrieval-Augmented Generation) for production insurance applications.


English

Model Overview

Item Description
Base model trillionlabs/Tri-7B
Fine-tuning method QLoRA / LoRA
Domain Insurance / Group Insurance / Claims Support
Primary language Korean
Model type Causal Language Model
Quantization GGUF F16 / Q4_K_M
Recommended deployment RAG-based insurance assistant
License Apache-2.0

Training Data

The model was fine-tuned using the insurance subset of a Korean financial customer-service dataset.

Initial insurance dataset:

  • Training: 24,000 QA samples
  • Validation: 3,000 QA samples

After removing automobile-insurance-focused samples and preparing the dataset for the intended group-insurance use case:

  • Training: 16,084 samples
  • Validation: 1,948 samples
  • Coverage / claims-related samples: approximately 70.5% of the training dataset

The training format was converted into a conversational structure:

System
  โ†“
Insurance assistant instructions

User
  โ†“
Instruction + customer question

Assistant
  โ†“
Reference answer

Fine-tuning

The primary QLoRA configuration included:

Base Model           trillionlabs/Tri-7B
Quantization         4-bit NF4
LoRA Rank            16
LoRA Alpha           32
LoRA Dropout         0.05

Target Modules
- q_proj
- k_proj
- v_proj
- o_proj

Max Sequence Length  512
Batch Size           1
Gradient Accumulation 8
Learning Rate        1e-4
Primary Epochs       1

More than 99.9% of the prepared training samples fit within 512 tokens.

Training Iterations

v1

The first model was trained directly on the prepared insurance consultation data.

It learned insurance terminology and consultation patterns, but it sometimes generated unsupported statements about:

  • Coverage eligibility
  • Claim payment eligibility
  • Enrolled benefits
  • Customer-specific policy information

v2

The dataset was modified to reduce unsupported conclusions when policy information was not available.

The model was trained to verify:

  • Actual coverage enrollment
  • Policy terms
  • Insurance period
  • Payment conditions
  • Exclusions

before making conclusions about claim eligibility.

v2.1 / v2.2

Additional calibration was performed for higher-risk insurance questions involving:

  • Required claim documents
  • Submission methods
  • Claim processing time
  • Field investigations
  • Claim amounts
  • Internal insurer procedures

These iterations improved conservative behavior, but they do not guarantee factual accuracy.

For this reason, production use should rely on RAG and deterministic safeguards.

Validation

An insurance-safety-adjusted validation set was also evaluated.

Example v2 results:

Training Loss:             1.7269

Original Validation Loss:  1.9097
Safe Validation Loss:      1.8079

Safe Validation
Mean Token Accuracy:       0.5542

These metrics measure language-model behavior on the prepared dataset and should not be interpreted as claim-decision accuracy.

Recommended Architecture

This model is designed to be used as part of the following architecture:

User Question
      โ†“
Risk / Intent Detection
      โ†“
RAG Retrieval
      โ†“
Customer Policy
Insurance Terms
Coverage Information
Claims Guide
      โ†“
Tri-7B Insurance v2.2
      โ†“
Grounded Response

The fine-tuned model should mainly provide:

  • Insurance-domain language understanding
  • Customer-service response style
  • Explanation of insurance terminology
  • Structured insurance consultation behavior

RAG should provide:

  • Actual policy terms
  • Customer-specific coverage
  • Benefit amounts
  • Required documents
  • Claim procedures
  • Submission channels
  • Current insurer rules

Important Limitations

This model must not be used as the sole source for insurance claim decisions.

The model may still generate incorrect or unsupported information, especially regarding:

  • Whether a claim is payable
  • Whether a specific benefit is enrolled
  • Required claim documents
  • Submission methods
  • Claim processing timelines
  • Benefit amounts
  • Insurer-specific procedures

For production systems, use:

Fine-tuned LLM
+
RAG
+
Policy / enrollment data
+
Output guardrails

If reliable source material is unavailable, the application should return a safe fallback response rather than allowing the model to guess.

Available Formats

This repository may contain:

Merged Hugging Face model
GGUF F16
GGUF Q4_K_M

For local inference with LM Studio or llama.cpp, the Q4_K_M GGUF version is recommended for a good balance between model quality and memory usage.

Disclaimer

This model is intended for research, prototyping, and insurance customer-support assistance.

It is not an automated claims adjudication system and should not independently determine coverage, liability, claim eligibility, or payment amounts.

Final insurance decisions must be based on the actual insurance contract, policy wording, enrollment information, applicable insurer rules, and qualified human review where required.


ํ•œ๊ตญ์–ด

๋ชจ๋ธ ๊ฐœ์š”

Tri-7B Insurance v2.2๋Š” trillionlabs/Tri-7B๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ๊ตญ์–ด ๋ณดํ—˜์ƒ๋‹ด ์—…๋ฌด์— ๋งž๊ฒŒ QLoRA ํŒŒ์ธํŠœ๋‹ํ•œ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.

์ฃผ์š” ํ•™์Šต ๋ชฉ์ ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค.

  • ๋ณดํ—˜ ์šฉ์–ด ์ดํ•ด ๋ฐ ์„ค๋ช…
  • ๊ธฐ์—… ๋‹จ์ฒด๋ณดํ—˜ ์ƒ๋‹ด
  • ๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ ๊ด€๋ จ ์งˆ์˜์‘๋‹ต
  • ๋ณด์žฅ๋‚ด์šฉ ์„ค๋ช…
  • ๊ฐ€์ž…๋‚ด์—ญ์ด๋‚˜ ์•ฝ๊ด€์ด ์—†๋Š” ๊ฒฝ์šฐ ์„ฑ๊ธ‰ํ•˜๊ฒŒ ์ง€๊ธ‰ ์—ฌ๋ถ€๋ฅผ ๋‹จ์ •ํ•˜์ง€ ์•Š๋Š” ์ƒ๋‹ด ๋ฐฉ์‹
  • ๊ณ ๊ฐ์ด ์ดํ•ดํ•˜๊ธฐ ์‰ฌ์šด ๋ณดํ—˜์ƒ๋‹ด ๋‹ต๋ณ€ ์ƒ์„ฑ

์‹ค์ œ ๋ณดํ—˜ ์„œ๋น„์Šค์—์„œ๋Š” RAG์™€ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์„ ์ „์ œ๋กœ ์„ค๊ณ„ํ–ˆ์Šต๋‹ˆ๋‹ค.

๊ธฐ๋ณธ ์ •๋ณด

ํ•ญ๋ชฉ ๋‚ด์šฉ
Base Model trillionlabs/Tri-7B
Fine-tuning QLoRA / LoRA
๋ถ„์•ผ ๋ณดํ—˜ / ๋‹จ์ฒด๋ณดํ—˜ / ๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ
์ฃผ์š” ์–ธ์–ด ํ•œ๊ตญ์–ด
๋ชจ๋ธ ์ข…๋ฅ˜ Causal Language Model
์ œ๊ณต ํ˜•์‹ Hugging Face / GGUF F16 / GGUF Q4_K_M
๊ถŒ์žฅ ์‚ฌ์šฉ ๋ฐฉ์‹ RAG ๊ธฐ๋ฐ˜ ๋ณดํ—˜์ƒ๋‹ด AI
๋ผ์ด์„ ์Šค Apache-2.0

ํ•™์Šต ๋ฐ์ดํ„ฐ

ํ•œ๊ตญ์–ด ๊ธˆ์œต ๊ณ ๊ฐ์ƒ๋‹ด ๋ฐ์ดํ„ฐ ์ค‘ ๋ณดํ—˜ ๋ถ„์•ผ ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ–ˆ์Šต๋‹ˆ๋‹ค.

์ตœ์ดˆ ๋ณดํ—˜ ๋ฐ์ดํ„ฐ:

Training     24,000๊ฑด
Validation    3,000๊ฑด

์ž๋™์ฐจ๋ณดํ—˜ ์ค‘์‹ฌ ๋ฐ์ดํ„ฐ๋ฅผ ์ œ๊ฑฐํ•˜๊ณ  ๊ธฐ์—… ๋‹จ์ฒด๋ณดํ—˜ยท๋ณด์žฅยท๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ ์—…๋ฌด์— ๋งž๊ฒŒ ์ •๋ฆฌํ•œ ๋ฐ์ดํ„ฐ๋Š”:

Training     16,084๊ฑด
Validation    1,948๊ฑด

๋ณด์žฅ/์ฒญ๊ตฌ ํ•ต์‹ฌ ๋ฐ์ดํ„ฐ ๋น„์œจ
์•ฝ 70.5%

์ž…๋‹ˆ๋‹ค.

ํ•™์Šต ๋ฐ์ดํ„ฐ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋Œ€ํ™” ํ˜•์‹์œผ๋กœ ๊ตฌ์„ฑํ–ˆ์Šต๋‹ˆ๋‹ค.

System
  โ†“
๋ณดํ—˜์ƒ๋‹ด AI ์—ญํ•  ๋ฐ ๊ทœ์น™

User
  โ†“
์ง€์‹œ๋ฌธ + ๊ณ ๊ฐ ์งˆ๋ฌธ

Assistant
  โ†“
๋ณดํ—˜์ƒ๋‹ด ๋‹ต๋ณ€

์ฃผ์š” QLoRA ์„ค์ •

Base Model           trillionlabs/Tri-7B

QLoRA                NF4 4bit

LoRA Rank            16
LoRA Alpha           32
LoRA Dropout         0.05

Target Modules
- q_proj
- k_proj
- v_proj
- o_proj

Max Length           512
Batch Size           1
Gradient Accumulation 8
Learning Rate        1e-4
Primary Epoch        1

์ „์ฒ˜๋ฆฌ๋œ Training ๋ฐ์ดํ„ฐ์˜ 99.9% ์ด์ƒ์ด 512 token ์ด๋‚ด์— ํฌํ•จ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

๋ชจ๋ธ ๊ฐœ์„  ๊ณผ์ •

V1

๋ณดํ—˜ ์ƒ๋‹ด ๋ฐ์ดํ„ฐ ์ž์ฒด๋ฅผ ์ด์šฉํ•ด ์ตœ์ดˆ ํŒŒ์ธํŠœ๋‹์„ ์ˆ˜ํ–‰ํ–ˆ์Šต๋‹ˆ๋‹ค.

๋ณดํ—˜ ์šฉ์–ด์™€ ์ƒ๋‹ด ๋ฌธ์žฅ์€ ์ž˜ ํ•™์Šตํ–ˆ์ง€๋งŒ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋ฌธ์ œ๊ฐ€ ๋ฐœ๊ฒฌ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

์‹ค์ œ ๊ฐ€์ž…๋‚ด์šฉ์„ ๋ชจ๋ฅด๋Š”๋ฐ
"ํ˜„์žฌ ๊ฐ€์ž…๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค"

์•ฝ๊ด€์„ ๋ชจ๋ฅด๋Š”๋ฐ
"๋ณดํ—˜๊ธˆ์ด ์ง€๊ธ‰๋ฉ๋‹ˆ๋‹ค"

๊ณ„์•ฝ์ •๋ณด๊ฐ€ ์—†๋Š”๋ฐ
"์ˆ˜์ˆ ๋น„ ๋‹ด๋ณด๊ฐ€ ์—†์Šต๋‹ˆ๋‹ค"

์™€ ๊ฐ™์ด ์กด์žฌํ•˜์ง€ ์•Š๋Š” ๊ณ ๊ฐ ๊ณ„์•ฝ์ •๋ณด๋ฅผ ๋งŒ๋“ค์–ด๋‚ด๋Š” ๋ฌธ์ œ๊ฐ€ ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค.

V2

๋ณด์žฅ ์—ฌ๋ถ€๋‚˜ ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰ ๊ฐ€๋Šฅ์„ฑ์„ ๋ฌป๋Š” ์งˆ๋ฌธ์—์„œ ์‹ค์ œ ๊ฐ€์ž…๋‚ด์—ญ์ด๋‚˜ ์•ฝ๊ด€์ด ์—†๋Š” ๊ฒฝ์šฐ ๋‹จ์ •ํ•˜์ง€ ์•Š๋„๋ก Training ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ •ํ–ˆ์Šต๋‹ˆ๋‹ค.

๋ชจ๋ธ์ด ๋‹ค์Œ ์‚ฌํ•ญ์„ ๋จผ์ € ํ™•์ธํ•˜๋„๋ก ํ•™์Šต ๋ฐฉํ–ฅ์„ ์กฐ์ •ํ–ˆ์Šต๋‹ˆ๋‹ค.

๊ฐ€์ž… ๋‹ด๋ณด
๋ณดํ—˜๊ธฐ๊ฐ„
์•ฝ๊ด€์ƒ ์ง€๊ธ‰์‚ฌ์œ 
๋ฉด์ฑ…์‚ฌํ•ญ
์‚ฌ๊ณ  ๋ฐ ์ง„๋‹จ ๋‚ด์šฉ

V2.1 / V2.2

์ถ”๊ฐ€์ ์œผ๋กœ ๋‹ค์Œ ์งˆ๋ฌธ ์œ ํ˜•์— ๋Œ€ํ•œ ๋ณด์ •ํ•™์Šต์„ ์ง„ํ–‰ํ–ˆ์Šต๋‹ˆ๋‹ค.

๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ์„œ๋ฅ˜
์„œ๋ฅ˜ ์ œ์ถœ๋ฐฉ๋ฒ•
๋ณดํ—˜๊ธˆ ์ฒ˜๋ฆฌ๊ธฐ๊ฐ„
ํ˜„์žฅ์กฐ์‚ฌ ์—ฌ๋ถ€
์ง€๊ธ‰๊ธˆ์•ก
๋ณดํ—˜์‚ฌ ๋‚ด๋ถ€ ์ฒ˜๋ฆฌ์ ˆ์ฐจ

์ด ๊ณผ์ •์„ ํ†ตํ•ด ๊ทผ๊ฑฐ ์—†๋Š” ๋‹จ์ •์€ ๊ฐ์†Œํ–ˆ์ง€๋งŒ Fine-tuning๋งŒ์œผ๋กœ ํ™˜๊ฐ์„ ์™„์ „ํžˆ ์ œ๊ฑฐํ•  ์ˆ˜๋Š” ์—†๋‹ค๋Š” ์ ๋„ ํ™•์ธํ–ˆ์Šต๋‹ˆ๋‹ค.

๋”ฐ๋ผ์„œ ์ตœ์ข… ์„œ๋น„์Šค์—์„œ๋Š” RAG์™€ Guard๋ฅผ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

Validation ๊ฒฐ๊ณผ

V2 ๋ชจ๋ธ์˜ ์ฃผ์š” ๊ฒฐ๊ณผ:

Train Loss
1.7269

๊ธฐ์กด Validation Loss
1.9097

์•ˆ์ „์„ฑ ์ˆ˜์ • Validation Loss
1.8079

์•ˆ์ „ Validation
Mean Token Accuracy
0.5542

์œ„ ์ˆ˜์น˜๋Š” ํ•™์Šต ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ Language Model ์ง€ํ‘œ์ด๋ฉฐ ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰ ํŒ๋‹จ ์ •ํ™•๋„๋ฅผ ์˜๋ฏธํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

๊ถŒ์žฅ ์„œ๋น„์Šค ๊ตฌ์กฐ

์‹ค์ œ ์„œ๋น„์Šค์—์„œ๋Š” ๋‹ค์Œ ๊ตฌ์กฐ๋ฅผ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

๊ณ ๊ฐ ์งˆ๋ฌธ
   โ†“
์งˆ๋ฌธ ์œ ํ˜• / ์œ„ํ—˜๋„ ํ™•์ธ
   โ†“
RAG ๊ฒ€์ƒ‰
   โ†“
๊ณ ๊ฐ์‚ฌ ๊ฐ€์ž…๋‚ด์—ญ
๋ณดํ—˜์•ฝ๊ด€
๋ณด์žฅ๋‚ด์—ญ
๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ์•ˆ๋‚ด
   โ†“
Tri-7B Insurance v2.2
   โ†“
๊ทผ๊ฑฐ ๊ธฐ๋ฐ˜ ๋‹ต๋ณ€

Fine-tuning ๋ชจ๋ธ์€ ์ฃผ๋กœ:

๋ณดํ—˜์—…๋ฌด ์–ธ์–ด ์ดํ•ด
๋ณดํ—˜์ƒ๋‹ด ๋ฌธ์ฒด
๋ณดํ—˜ ์šฉ์–ด ์„ค๋ช…
๋ณดํ—˜ ์งˆ๋ฌธ ์˜๋„ ์ดํ•ด
์ƒ๋‹ด ๋‹ต๋ณ€ ๊ตฌ์„ฑ

์„ ๋‹ด๋‹นํ•ฉ๋‹ˆ๋‹ค.

RAG๋Š”:

์‹ค์ œ ๊ฐ€์ž… ๋‹ด๋ณด
๋ณดํ—˜๊ฐ€์ž…๊ธˆ์•ก
๊ณ ๊ฐ์‚ฌ๋ณ„ ๋ณด์žฅ๋‚ด์šฉ
๋ณดํ—˜์•ฝ๊ด€
์ฒญ๊ตฌ์„œ๋ฅ˜
๋ณดํ—˜๊ธˆ ์ ‘์ˆ˜๋ฐฉ๋ฒ•
๋ณดํ—˜์‚ฌ๋ณ„ ์ตœ์‹  ๊ธฐ์ค€

๊ณผ ๊ฐ™์€ ์‚ฌ์‹ค์ •๋ณด๋ฅผ ๋‹ด๋‹นํ•˜๋„๋ก ์„ค๊ณ„ํ•˜๋Š” ๊ฒƒ์ด ์ข‹์Šต๋‹ˆ๋‹ค.

์ค‘์š”ํ•œ ์ œํ•œ์‚ฌํ•ญ

๋ณธ ๋ชจ๋ธ์˜ ์ถœ๋ ฅ๋งŒ์œผ๋กœ ๋‹ค์Œ ์‚ฌํ•ญ์„ ๊ฒฐ์ •ํ•ด์„œ๋Š” ์•ˆ ๋ฉ๋‹ˆ๋‹ค.

  • ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰ ์—ฌ๋ถ€
  • ์‹ค์ œ ๋‹ด๋ณด ๊ฐ€์ž… ์—ฌ๋ถ€
  • ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰๊ธˆ์•ก
  • ๋ฉด์ฑ… ์—ฌ๋ถ€
  • ๋ณดํ—˜๊ธˆ ์ฒญ๊ตฌ ํ•„์ˆ˜์„œ๋ฅ˜
  • ๋ณดํ—˜๊ธˆ ์ ‘์ˆ˜๋ฐฉ๋ฒ•
  • ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰๊ธฐ๊ฐ„
  • ๋ณดํ—˜์‚ฌ ๋‚ด๋ถ€ ์ฒ˜๋ฆฌ์ ˆ์ฐจ

๋ชจ๋ธ์€ ์ด๋Ÿฌํ•œ ๋‚ด์šฉ์„ ์ž˜๋ชป ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์‹ค์ œ ๋ณดํ—˜ ์„œ๋น„์Šค์—์„œ๋Š” ๋‹ค์Œ ๊ตฌ์กฐ๋ฅผ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

Fine-tuned LLM
+
RAG
+
๊ณ ๊ฐ์‚ฌ ๊ฐ€์ž…์ •๋ณด
+
๋ณดํ—˜์•ฝ๊ด€
+
Guardrail

๊ด€๋ จ ๊ทผ๊ฑฐ์ž๋ฃŒ๊ฐ€ ๊ฒ€์ƒ‰๋˜์ง€ ์•Š์€ ๊ฒฝ์šฐ์—๋Š” ๋ชจ๋ธ์ด ์ž„์˜๋กœ ๋‹ต๋ณ€ํ•˜๋„๋ก ํ•˜์ง€ ์•Š๊ณ , ์•ˆ์ „ํ•œ ์•ˆ๋‚ด๋ฌธ์„ ๋ฐ˜ํ™˜ํ•˜๋Š” ๋ฐฉ์‹์ด ์ ์ ˆํ•ฉ๋‹ˆ๋‹ค.

์ œ๊ณต ๋ชจ๋ธ ํ˜•์‹

Repository์—๋Š” ๋‹ค์Œ ํ˜•์‹์„ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

1. Hugging Face Merged Model

2. GGUF F16

3. GGUF Q4_K_M

LM Studio ๋˜๋Š” llama.cpp์—์„œ ๋กœ์ปฌ ์‹คํ–‰ํ•˜๋Š” ๊ฒฝ์šฐ์—๋Š” Q4_K_M ๋ฒ„์ „์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.

๋ฉด์ฑ… ๋ฐ ์‚ฌ์šฉ์ƒ ์ฃผ์˜

๋ณธ ๋ชจ๋ธ์€ ๋ณดํ—˜ AI ์—ฐ๊ตฌ, ํ”„๋กœํ† ํƒ€์ž… ๋ฐ ๋ณดํ—˜์ƒ๋‹ด ์—…๋ฌด ์ง€์›์„ ๋ชฉ์ ์œผ๋กœ ์ œ์ž‘๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

๋ณธ ๋ชจ๋ธ์€ ์ž๋™ ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰์‹ฌ์‚ฌ ๋ชจ๋ธ์ด ์•„๋‹™๋‹ˆ๋‹ค.

๋ชจ๋ธ์˜ ์ถœ๋ ฅ๋งŒ์œผ๋กœ ๋ณด์žฅ ์—ฌ๋ถ€, ๋ณดํ—˜๊ธˆ ์ง€๊ธ‰ ์—ฌ๋ถ€, ์ฑ…์ž„ ์—ฌ๋ถ€ ๋˜๋Š” ์ง€๊ธ‰๊ธˆ์•ก์„ ๊ฒฐ์ •ํ•ด์„œ๋Š” ์•ˆ ๋ฉ๋‹ˆ๋‹ค.

์ตœ์ข… ํŒ๋‹จ์€ ๋ฐ˜๋“œ์‹œ ์‹ค์ œ ๋ณดํ—˜๊ณ„์•ฝ, ๋ณดํ—˜์•ฝ๊ด€, ๊ฐ€์ž…๋‚ด์—ญ, ๋ณดํ—˜์‚ฌ ๊ธฐ์ค€ ๋ฐ ํ•„์š”ํ•œ ๊ฒฝ์šฐ ๋‹ด๋‹น์ž์˜ ๊ฒ€ํ† ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์ด๋ฃจ์–ด์ ธ์•ผ ํ•ฉ๋‹ˆ๋‹ค.

Downloads last month
84
Safetensors
Model size
8B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for KokAI-lab/Tri-7B-Insurance-v22

Adapter
(2)
this model