ThingsAI commited on
Commit
f758f7b
·
verified ·
1 Parent(s): bf2db5d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -2
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 (pretrained on 15 B tokens of general‑purpose and mathematical text)
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.