Update README.md
Browse files
README.md
CHANGED
|
@@ -7,4 +7,47 @@ sdk: static
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
| 10 |
+
# DeepXR
|
| 11 |
+
|
| 12 |
+
**Building Intelligent, Responsible, and Accessible AI Systems**
|
| 13 |
+
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
## About Us
|
| 17 |
+
|
| 18 |
+
DeepXR is an AI research organization dedicated to developing state-of-the-art language models that are powerful, safe, and accessible. We advance AI technology while maintaining the highest standards of safety, ethics, and transparency.
|
| 19 |
+
|
| 20 |
+
## Our Mission
|
| 21 |
+
|
| 22 |
+
To democratize access to advanced AI capabilities by developing open-source models that balance performance, safety, and efficiency—enabling developers and researchers worldwide to build transformative applications.
|
| 23 |
+
|
| 24 |
+
## Core Values
|
| 25 |
+
|
| 26 |
+
**Innovation** • Pushing boundaries through cutting-edge research and engineering
|
| 27 |
+
|
| 28 |
+
**Safety First** • Comprehensive safety measures and ethical guidelines in every release
|
| 29 |
+
|
| 30 |
+
**Openness** • Contributing to open-source with transparent documentation and accessible models
|
| 31 |
+
|
| 32 |
+
**Responsibility** • Developing AI with consideration for societal impact and user wellbeing
|
| 33 |
+
|
| 34 |
+
## Our Models
|
| 35 |
+
|
| 36 |
+
### Helion Series
|
| 37 |
+
Production-ready language models designed for real-world applications:
|
| 38 |
+
|
| 39 |
+
- **Helion-V2 (7.2B)**: Advanced LLM with exceptional truthfulness and balanced capabilities
|
| 40 |
+
- Strong performance across reasoning, coding, and conversational tasks
|
| 41 |
+
- Built-in safety features and content moderation
|
| 42 |
+
- Optimized for efficient deployment on consumer hardware
|
| 43 |
+
|
| 44 |
+
## Key Achievements
|
| 45 |
+
|
| 46 |
+
- Leading TruthfulQA scores in parameter class (52.1%)
|
| 47 |
+
- Industry-leading safety metrics (0.08 ToxiGen)
|
| 48 |
+
- Production-ready with real-world deployments
|
| 49 |
+
- Apache 2.0 licensed for research and commercial use
|
| 50 |
+
|
| 51 |
+
---
|
| 52 |
+
|
| 53 |
+
*DeepXR • Advancing AI for Everyone • 2024*
|