Karte: explizite Farben für Überschriften und Fettdruck
Browse files
README.md
CHANGED
|
@@ -7,11 +7,13 @@ sdk: static
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
|
|
|
|
|
|
| 10 |

|
| 11 |
|
| 12 |
-
Keyvan.ai is an independent AI research and engineering practice based in Germany (Keyvan Hardani). We build and evaluate language, vision, audio and security models for production use, with a focus on small models that run on-premises, and we publish our experiments with full protocols.
|
| 13 |
|
| 14 |
-
#
|
| 15 |
|
| 16 |
Editing the weights of neural networks directly with a frontier model as the agent, without gradients or training, verified by a fixed evaluation harness. Following the MIT CSAIL study *Sparks of Weight-Space Programming* (2 November 2026), we replicated and extended the approach with Claude Mythos 5.1.
|
| 17 |
|
|
@@ -21,32 +23,34 @@ Editing the weights of neural networks directly with a frontier model as the age
|
|
| 21 |
|
| 22 |

|
| 23 |
|
| 24 |
-
#
|
| 25 |
|
| 26 |
-
|
| 27 |
- [Mankei-1B-Chat](https://huggingface.co/keyvan-ai/Mankei-1B-Chat) · [Mankei-1B-Chat-GGUF](https://huggingface.co/keyvan-ai/Mankei-1B-Chat-GGUF) – 1B German chat model, safetensors and GGUF
|
| 28 |
- [Mankei-Boarisch](https://huggingface.co/keyvan-ai/Mankei-Boarisch) – 1B Bavarian-dialect text generation
|
| 29 |
- [Mankei-326M-Embedder](https://huggingface.co/keyvan-ai/Mankei-326M-Embedder) · [Mankei-326M-Reranker](https://huggingface.co/keyvan-ai/Mankei-326M-Reranker) – German retrieval: sentence embeddings and reranking
|
| 30 |
- [Mankei-Reflex](https://huggingface.co/keyvan-ai/Mankei-Reflex) – text classification
|
| 31 |
- [german-text-3.1](https://huggingface.co/keyvan-ai/german-text-3.1) – 2B German text generation
|
| 32 |
|
| 33 |
-
|
| 34 |
- [german-ocr-3.1](https://huggingface.co/keyvan-ai/german-ocr-3.1) (2B) · [german-ocr-3](https://huggingface.co/keyvan-ai/german-ocr-3) (0.8B) – image-text-to-text models for German documents
|
| 35 |
- [german-ocr](https://huggingface.co/keyvan-ai/german-ocr) · [german-ocr-2b-gguf](https://huggingface.co/keyvan-ai/german-ocr-2b-gguf) · [german-ocr-turbo](https://huggingface.co/keyvan-ai/german-ocr-turbo) · [g-ocr](https://huggingface.co/keyvan-ai/g-ocr) – earlier generations and GGUF builds
|
| 36 |
|
| 37 |
-
|
| 38 |
- [belegant-4b](https://huggingface.co/keyvan-ai/belegant-4b) – 4B image-text-to-text model for receipts and business documents
|
| 39 |
- [AnomalyDetection-MVTech-Metal](https://huggingface.co/keyvan-ai/AnomalyDetection-MVTech-Metal) – image segmentation for surface anomalies on metal parts
|
| 40 |
|
| 41 |
-
|
| 42 |
- [wakecore](https://huggingface.co/keyvan-ai/wakecore) – wake-word and audio classification
|
| 43 |
|
| 44 |
-
|
| 45 |
- [CanDefender](https://huggingface.co/keyvan-ai/CanDefender) · [CanDefender-Fuzzy](https://huggingface.co/keyvan-ai/CanDefender-Fuzzy) · [CANDefender-DoS](https://huggingface.co/keyvan-ai/CANDefender-DoS) – CAN-bus intrusion detection (fuzzing and DoS attacks)
|
| 46 |
- [SecIDS-v2](https://huggingface.co/keyvan-ai/SecIDS-v2) · [SecIDS-CNN](https://huggingface.co/keyvan-ai/SecIDS-CNN) – network intrusion detection
|
| 47 |
|
| 48 |
-
#
|
| 49 |
|
| 50 |
Every published result comes with the protocol it was measured under, the controls that were run, and the limitations we found. Models are released under their base-model licences; reports under CC BY 4.0.
|
| 51 |
|
| 52 |
[keyvan.ai](https://keyvan.ai) · [GitHub](https://github.com/keyvanhardani) · [LinkedIn](https://www.linkedin.com/in/keyvanhardani)
|
|
|
|
|
|
|
|
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
|
| 10 |
+
<div style="background:#0b0f1a;color:#dbe4ee;padding:24px 28px;border-radius:12px">
|
| 11 |
+
|
| 12 |

|
| 13 |
|
| 14 |
+
<strong style="color:#ffffff">Keyvan.ai</strong> is an independent AI research and engineering practice based in Germany (Keyvan Hardani). We build and evaluate language, vision, audio and security models for production use, with a focus on small models that run on-premises, and we publish our experiments with full protocols.
|
| 15 |
|
| 16 |
+
<h2 style="color:#ffffff;font-size:1.25em;margin:1.2em 0 .4em">Current research: weight-space programming</h2>
|
| 17 |
|
| 18 |
Editing the weights of neural networks directly with a frontier model as the agent, without gradients or training, verified by a fixed evaluation harness. Following the MIT CSAIL study *Sparks of Weight-Space Programming* (2 November 2026), we replicated and extended the approach with Claude Mythos 5.1.
|
| 19 |
|
|
|
|
| 23 |
|
| 24 |

|
| 25 |
|
| 26 |
+
<h2 style="color:#ffffff;font-size:1.25em;margin:1.2em 0 .4em">Models</h2>
|
| 27 |
|
| 28 |
+
<p style="color:#ffffff;font-weight:700;margin:1em 0 .2em">German language (Mankei family)</p>
|
| 29 |
- [Mankei-1B-Chat](https://huggingface.co/keyvan-ai/Mankei-1B-Chat) · [Mankei-1B-Chat-GGUF](https://huggingface.co/keyvan-ai/Mankei-1B-Chat-GGUF) – 1B German chat model, safetensors and GGUF
|
| 30 |
- [Mankei-Boarisch](https://huggingface.co/keyvan-ai/Mankei-Boarisch) – 1B Bavarian-dialect text generation
|
| 31 |
- [Mankei-326M-Embedder](https://huggingface.co/keyvan-ai/Mankei-326M-Embedder) · [Mankei-326M-Reranker](https://huggingface.co/keyvan-ai/Mankei-326M-Reranker) – German retrieval: sentence embeddings and reranking
|
| 32 |
- [Mankei-Reflex](https://huggingface.co/keyvan-ai/Mankei-Reflex) – text classification
|
| 33 |
- [german-text-3.1](https://huggingface.co/keyvan-ai/german-text-3.1) – 2B German text generation
|
| 34 |
|
| 35 |
+
<p style="color:#ffffff;font-weight:700;margin:1em 0 .2em">German document understanding (German-OCR family)</p>
|
| 36 |
- [german-ocr-3.1](https://huggingface.co/keyvan-ai/german-ocr-3.1) (2B) · [german-ocr-3](https://huggingface.co/keyvan-ai/german-ocr-3) (0.8B) – image-text-to-text models for German documents
|
| 37 |
- [german-ocr](https://huggingface.co/keyvan-ai/german-ocr) · [german-ocr-2b-gguf](https://huggingface.co/keyvan-ai/german-ocr-2b-gguf) · [german-ocr-turbo](https://huggingface.co/keyvan-ai/german-ocr-turbo) · [g-ocr](https://huggingface.co/keyvan-ai/g-ocr) – earlier generations and GGUF builds
|
| 38 |
|
| 39 |
+
<p style="color:#ffffff;font-weight:700;margin:1em 0 .2em">Vision and industrial inspection</p>
|
| 40 |
- [belegant-4b](https://huggingface.co/keyvan-ai/belegant-4b) – 4B image-text-to-text model for receipts and business documents
|
| 41 |
- [AnomalyDetection-MVTech-Metal](https://huggingface.co/keyvan-ai/AnomalyDetection-MVTech-Metal) – image segmentation for surface anomalies on metal parts
|
| 42 |
|
| 43 |
+
<p style="color:#ffffff;font-weight:700;margin:1em 0 .2em">Audio</p>
|
| 44 |
- [wakecore](https://huggingface.co/keyvan-ai/wakecore) – wake-word and audio classification
|
| 45 |
|
| 46 |
+
<p style="color:#ffffff;font-weight:700;margin:1em 0 .2em">Security: intrusion detection for vehicle and IT networks</p>
|
| 47 |
- [CanDefender](https://huggingface.co/keyvan-ai/CanDefender) · [CanDefender-Fuzzy](https://huggingface.co/keyvan-ai/CanDefender-Fuzzy) · [CANDefender-DoS](https://huggingface.co/keyvan-ai/CANDefender-DoS) – CAN-bus intrusion detection (fuzzing and DoS attacks)
|
| 48 |
- [SecIDS-v2](https://huggingface.co/keyvan-ai/SecIDS-v2) · [SecIDS-CNN](https://huggingface.co/keyvan-ai/SecIDS-CNN) – network intrusion detection
|
| 49 |
|
| 50 |
+
<h2 style="color:#ffffff;font-size:1.25em;margin:1.2em 0 .4em">Principles</h2>
|
| 51 |
|
| 52 |
Every published result comes with the protocol it was measured under, the controls that were run, and the limitations we found. Models are released under their base-model licences; reports under CC BY 4.0.
|
| 53 |
|
| 54 |
[keyvan.ai](https://keyvan.ai) · [GitHub](https://github.com/keyvanhardani) · [LinkedIn](https://www.linkedin.com/in/keyvanhardani)
|
| 55 |
+
|
| 56 |
+
</div>
|