Update README.md
Browse files
README.md
CHANGED
|
@@ -16,7 +16,7 @@ tags:
|
|
| 16 |
|
| 17 |
Quark‑135M is a **135M parameter** conversational AI assistant, trained from scratch and then fine‑tuned to be **helpful, respectful, honest** and to remember a clear identity.
|
| 18 |
|
| 19 |
-
* **Base model:** Quark‑135M
|
| 20 |
* **Instruction tuning:** supervised fine‑tuning on a small, curated dataset of identity‑aware conversations
|
| 21 |
* **Developers:** OvercastLab and ThingsAI
|
| 22 |
* **License:** Apache‑2.0
|
|
@@ -90,7 +90,6 @@ It is **not suitable** for tasks requiring factual accuracy, deep reasoning, or
|
|
| 90 |
## Limitations
|
| 91 |
|
| 92 |
* **Small model size** – 135M parameters are an order of magnitude smaller than current frontier models.
|
| 93 |
-
* **Limited world knowledge** – pre‑trained on only 15 B tokens; it lacks the broad coverage of larger models.
|
| 94 |
* **Hallucinates frequently** – when asked questions beyond simple greetings or self‑description, it may invent plausible‑sounding but incorrect answers.
|
| 95 |
* **Repetitive loops** – may occasionally repeat phrases or get stuck in loops, especially with low temperature sampling.
|
| 96 |
* **Instruction coverage** – fine‑tuned on only 1 500 identity examples; it may not handle out‑of‑domain requests gracefully.
|
|
|
|
| 16 |
|
| 17 |
Quark‑135M is a **135M parameter** conversational AI assistant, trained from scratch and then fine‑tuned to be **helpful, respectful, honest** and to remember a clear identity.
|
| 18 |
|
| 19 |
+
* **Base model:** Quark‑135M
|
| 20 |
* **Instruction tuning:** supervised fine‑tuning on a small, curated dataset of identity‑aware conversations
|
| 21 |
* **Developers:** OvercastLab and ThingsAI
|
| 22 |
* **License:** Apache‑2.0
|
|
|
|
| 90 |
## Limitations
|
| 91 |
|
| 92 |
* **Small model size** – 135M parameters are an order of magnitude smaller than current frontier models.
|
|
|
|
| 93 |
* **Hallucinates frequently** – when asked questions beyond simple greetings or self‑description, it may invent plausible‑sounding but incorrect answers.
|
| 94 |
* **Repetitive loops** – may occasionally repeat phrases or get stuck in loops, especially with low temperature sampling.
|
| 95 |
* **Instruction coverage** – fine‑tuned on only 1 500 identity examples; it may not handle out‑of‑domain requests gracefully.
|