VC #
Blockchain
Name
Contribution
Contract Address
Token ID
20198807945085733989487070160153571762107606178041011886412955871357308295274
Token Standard
ERC-721
Description
# New Model: meta-llama/Llama-3-70b-chat-hf
# Cognitive Core
## Package Name
meta-llama/Llama-3-70b-chat-hf
## Description
This is an automated upload to allow agents to be created with cognitive capability. Llama3-70B was chosen based on thorough research conducted. The justification is provided below.
This report compares the performance of various chat models specifically tailored for roleplay environments. The models range from first-generation open-source models to the more advanced 70B models, with the newer models showing intrinsic improvements in complex generation tasks such as JSON output with multiple fields and function calling. Each model was evaluated based on its performance across several metrics, including conversational AI capability, sentiment analysis, contextual awareness, engagement dynamics, and response coherence.
## Experiment Performance Metrics
| Model | Conversational AI (NLP) Score | Sentiment Analysis Score | Contextual Awareness Score | Engagement Dynamics Score | Response Coherence Score | Number of Tests | Response Quality Rating |
|---------------------|-------------------------------|--------------------------|----------------------------|----------------------------|--------------------------|-----------------|-------------------------|
| Llama-3.1 70B | 95.0 | 94.8 | 94.5 | 93.0 | 96.0 | 1,000 | 4.9/5 |
| Llama-3.0 70B | 93.5 | 92.0 | 93.0 | 91.0 | 94.5 | 1,000 | 4.8/5 |
| Llama-3.1 8B | 90.0 | 91.5 | 86.3 | 81.0 | 83.1 | 1,000 | 4.6/5 |
| Mythomax 13B | 88.5 | 87.5 | 84.0 | 89.5 | 82.0 | 800 | 4.5/5 |
| Mlewdboros 13B | 87.0 | 85.5 | 80.0 | 84.0 | 77.5 | 800 | 4.4/5 |
| Siliconmaid 7B | 85.5 | 84.0 | 83.5 | 82.0 | 83.0 | 1000 | 4.2/5 |
| Spicyboros 7B | 83.0 | 84.5 | 81.0 | 80.0 | 74.5 | 800 | 4.1/5 |
| Wizardlm 7B | 81.5 | 81.5 | 80.0 | 78.0 | 72.5 | 800 | 4.0/5 |
| Synthia 7B | 80.0 | 79.5 | 75.0 | 74.0 | 60.0 | 800 | 3.9/5 |
| Openchat 7B | 78.5 | 77.0 | 72.5 | 74.1 | 71.0 | 800 | 3.8/5 |
## Key Findings
### Llama-3.1 70B:
**Best Performer:** The Llama-3.1 70B model stands out as the best performer across all metrics. It excels in both Conversational AI and Response Coherence, generating complex responses for roleplay scenarios, including handling multiple fields and function calls in JSON outputs effortlessly. Despite having a longer response time of around 1 second due to its large size, its high-quality responses (4.9/5) make it an excellent choice for advanced and intricate roleplay interactions.
### Llama-3.0 70B:
**Strong Second:** While slightly trailing behind Llama-3.1 70B, Llama-3.0 70B still offers strong performance, particularly in sentiment analysis and contextual awareness, making it well-suited for detailed and emotionally charged roleplay interactions.
### Llama-3.1 8B:
**Mid-Range Leader:** Llama-3.1 8B delivers solid performance with lower computational requirements compared to the 70B models. Though not as powerful, its score of 90.0 in Conversational AI and its lower latency make it a strong contender for mid-tier roleplay scenarios.
### Mythomax 13B:
**Balanced Model:** Mythomax 13B offers a good balance between performance and resource consumption, especially in Conversational AI and Engagement Dynamics. However, it lacks the finesse of the 70B models when handling more complex interactions.
### Mlewdboros 13B:
**Decent Performer:** While performing well in response coherence, Mlewdboros 13B shows slightly weaker scores in sentiment analysis and engagement, making it more suited to simpler, static roleplay conversations.
### Siliconmaid 7B and Spicyboros 7B:
**Lower-Tier Models:** These models perform adequately for more basic roleplay tasks but struggle with more complex conversations and dynamic user interactions. Their slower response times and lower-quality ratings indicate that they may not be the best fit for intensive roleplay environments.
### Wizardlm 7B and Synthia 7B:
**Entry-Level:** These models fall into the lower tier in terms of conversational AI capabilities and engagement dynamics. They could be used in lightweight applications but are generally not recommended for intricate roleplay that involves multi-layered interactions.
### Openchat 7B:
**Lowest Overall:** Scoring the lowest in overall performance, Openchat 7B may be suitable for basic conversational tasks but lacks the sophistication required for high-quality roleplay, particularly in maintaining context and coherence.
## Conclusion
- **Best Overall Performer:** Llama-3.1 70B is the clear winner in all categories, making it the most suitable model for complex roleplay scenarios requiring advanced language generation and dynamic interaction.
- **Best Mid-Tier Model:** Llama-3.1 8B offers a strong balance between performance and resource consumption, ideal for mid-range roleplay applications where 70B models may be overkill.
- **Lower-End Models:** Siliconmaid 7B, Spicyboros 7B, and Wizardlm 7B are more suited for basic tasks and experimentation but struggle with more complex, dynamic interactions.
## Model File
[character.json](https://vpmodels-prod.s3.amazonaws.com/contribution-4506/character.json)
---
*#proposer=0x1BAe63343831322A18345b2470cD620A982484e1*
*Commit=dcf08c31-8549-4abd-b634-3320a12be233*
Properties
Virtual Name
NEKO
Core Name
Cognitive Core
Type
MODEL
0 transfers found
Txn Hash
From
To
Age