CollectiveCognition-v1.1-Mistral-7B Open-source Dialog Model - Free Deployment, Excellent Performance in Q&A Tests

Home

Collectivecognition V1.1 Mistral 7B

Developed by teknium

An efficient dialogue model fine-tuned based on Mistral-7B, excelling in TruthfulQA benchmark tests

Large Language Model

Transformers

EnglishOpen Source License:Apache-2.0 #Efficient Fine-tuning #TruthfulQA Leader #Few-shot Optimization

Downloads 63

Release Time : 10/4/2023

Model Overview

An advanced dialogue model fine-tuned using the Mistral approach, particularly adept at identifying and correcting common misconceptions, performing on par with 70B-scale models in TruthfulQA benchmarks

Model Features

Efficient Training

Training completed in just 3 minutes on a single 4090 GPU using qlora

Few-shot Efficiency

Achieves outstanding performance with only 100 sample data points for training

Excellent TruthfulQA Performance

Surpasses multiple 70B-scale models in TruthfulQA benchmark tests

Low-cost Deployment

7B parameter size makes it suitable for deployment in resource-limited environments

Model Capabilities

Dialogue Generation

Question Answering System

Misconception Identification

Knowledge Correction

Use Cases

Education

Knowledge Q&A System

Used for intelligent question answering in educational settings

Capable of accurately identifying and correcting common knowledge misconceptions

Customer Service

Intelligent Customer Service Dialogue

Provides efficient customer service dialogue support

🚀 Collective Cognition v1.1 - Mistral 7B

Collective Cognition v1.1 is a state - of - the - art model fine - tuned using the Mistral approach, outperforming many 70B models on the TruthfulQA benchmark.

📚 Documentation

Model Description

Collective Cognition v1.1 is a state - of - the - art model fine - tuned using the Mistral approach. This model is particularly notable for its performance, outperforming many 70B models on the TruthfulQA benchmark. This benchmark assesses models for common misconceptions, potentially indicating hallucination rates.

✨ Features

Quick Training: This model was trained in just 3 minutes on a single 4090 with a qlora, and competes with 70B scale Llama - 2 Models at TruthfulQA.
Limited Data: Despite its exceptional performance, it was trained on only ONE HUNDRED data points, all of which were gathered from a platform reminiscent of ShareGPT.
Extreme TruthfulQA Benchmark: This model is competing strongly with top 70B models on the TruthfulQA benchmark despite the small dataset and qlora training!

image/png

Acknowledgements

Special thanks to @a16z and all contributors to the Collective Cognition dataset for making the development of this model possible.

Dataset

The model was trained using data from the Collective Cognition website. The efficacy of this dataset is demonstrated by the model's stellar performance, suggesting that further expansion of this dataset could yield even more promising results. The data is reminiscent of that collected from platforms like ShareGPT.

You can contribute to the growth of the dataset by sharing your own ChatGPT chats here.

You can download the datasets created by Collective Cognition here: https://huggingface.co/CollectiveCognition

Performance

TruthfulQA: Collective Cognition v1.1 has notably outperformed various 70B models on the TruthfulQA benchmark, highlighting its ability to understand and rectify common misconceptions.

💻 Usage Examples

Basic Usage

Prompt Format:

USER: <prompt>
ASSISTANT:

<system message>
USER: <prompt>
ASSISTANT:

Benchmarks

Collective Cognition v1.0 TruthfulQA:

|    Task     |Version|Metric|Value |   |Stderr|
|-------------|------:|------|-----:|---|-----:|
|truthfulqa_mc|      1|mc1   |0.4051|±  |0.0172|
|             |       |mc2   |0.5738|±  |0.0157|

Collective Cognition v1.1 GPT4All:

|    Task     |Version| Metric |Value |   |Stderr|
|-------------|------:|--------|-----:|---|-----:|
|arc_challenge|      0|acc     |0.5085|±  |0.0146|
|             |       |acc_norm|0.5384|±  |0.0146|
|arc_easy     |      0|acc     |0.7963|±  |0.0083|
|             |       |acc_norm|0.7668|±  |0.0087|
|boolq        |      1|acc     |0.8495|±  |0.0063|
|hellaswag    |      0|acc     |0.6399|±  |0.0048|
|             |       |acc_norm|0.8247|±  |0.0038|
|openbookqa   |      0|acc     |0.3240|±  |0.0210|
|             |       |acc_norm|0.4540|±  |0.0223|
|piqa         |      0|acc     |0.7992|±  |0.0093|
|             |       |acc_norm|0.8107|±  |0.0091|
|winogrande   |      0|acc     |0.7348|±  |0.0124|
Average: 71.13

AGIEval:

|             Task             |Version| Metric |Value |   |Stderr|
|------------------------------|------:|--------|-----:|---|-----:|
|agieval_aqua_rat              |      0|acc     |0.1929|±  |0.0248|
|                              |       |acc_norm|0.2008|±  |0.0252|
|agieval_logiqa_en             |      0|acc     |0.3134|±  |0.0182|
|                              |       |acc_norm|0.3333|±  |0.0185|
|agieval_lsat_ar               |      0|acc     |0.2217|±  |0.0275|
|                              |       |acc_norm|0.2043|±  |0.0266|
|agieval_lsat_lr               |      0|acc     |0.3412|±  |0.0210|
|                              |       |acc_norm|0.3216|±  |0.0207|
|agieval_lsat_rc               |      0|acc     |0.4721|±  |0.0305|
|                              |       |acc_norm|0.4201|±  |0.0301|
|agieval_sat_en                |      0|acc     |0.6068|±  |0.0341|
|                              |       |acc_norm|0.5777|±  |0.0345|
|agieval_sat_en_without_passage|      0|acc     |0.3932|±  |0.0341|
|                              |       |acc_norm|0.3641|±  |0.0336|
|agieval_sat_math              |      0|acc     |0.2864|±  |0.0305|
|                              |       |acc_norm|0.2636|±  |0.0298|
Average: 33.57

Training run on wandb here: https://wandb.ai/teknium1/collectivecognition-mistral-7b/runs/collectivecognition-mistral-8/workspace

📄 License

Apache 2.0

Property	Details
Base Model	mistralai/Mistral - 7B - v0.1
Tags	mistral - 7b, instruct, finetune, gpt4, synthetic data, distillation, sharegpt
Datasets	CollectiveCognition/chats - data - 2023 - 09 - 27
Model Name	CollectiveCognition - v1 - Mistral - 7B
License	apache - 2.0
Language	en

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご