OzTianlu commited on
Commit
d414dde
·
verified ·
1 Parent(s): e6277d9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +20 -18
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  library_name: transformers
3
- model_name: Asterisk-135M
4
  base_model: HuggingFaceTB/SmolLM2-135M-Instruct
5
  tags:
6
  - aspp
@@ -25,6 +25,24 @@ language:
25
  - **Training**: Supervised Fine-Tuning on Capybara dataset
26
  - **Framework**: Transformers 4.57.6, TRL 0.27.0
27
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28
  ### Key Innovation: The Asterisk Operator (★-operator)
29
 
30
  The **Asterisk Operator** performs local parallel state evolution through point-wise transformations:
@@ -152,23 +170,6 @@ class AsteriskForCausalLM(LlamaForCausalLM):
152
  """
153
  ```
154
 
155
- ## Evaluation Results
156
-
157
- Evaluated on LM-Evaluation-Harness with `limit=50` per task:
158
-
159
- | Task | Metric | Score | Stderr |
160
- |------|--------|-------|--------|
161
- | **MMLU** | acc | **0.2376** | ±0.0037 |
162
- | - Humanities | acc | 0.2472 | ±0.0067 |
163
- | - STEM | acc | 0.2245 | ±0.0074 |
164
- | - Social Sciences | acc | 0.2327 | ±0.0076 |
165
- | - Other | acc | 0.2430 | ±0.0077 |
166
- | **GSM8K** | exact_match | **0.0240** | ±0.0048 |
167
- | **HellaSwag** | acc_norm | **0.4430** | ±0.0157 |
168
- | **ARC-Easy** | acc_norm | **0.5450** | ±0.0158 |
169
- | **PIQA** | acc_norm | **0.6770** | ±0.0148 |
170
- | **WinoGrande** | acc | **0.5210** | ±0.0158 |
171
-
172
  **Note**: These are preliminary results with sample limits. Full evaluation pending.
173
 
174
  ## Quick Start
@@ -285,6 +286,7 @@ If you use this model, please cite:
285
  author={NoesisLab},
286
  year={2026},
287
  publisher={Huggingface},
 
288
  }
289
  ```
290
 
 
1
  ---
2
  library_name: transformers
3
+ model_name: Asterisk
4
  base_model: HuggingFaceTB/SmolLM2-135M-Instruct
5
  tags:
6
  - aspp
 
25
  - **Training**: Supervised Fine-Tuning on Capybara dataset
26
  - **Framework**: Transformers 4.57.6, TRL 0.27.0
27
 
28
+
29
+ ## Evaluation Results
30
+
31
+ Evaluated on LM-Evaluation-Harness:
32
+
33
+ | Task | Metric | Score | Stderr |
34
+ |------|--------|-------|--------|
35
+ | **MMLU** | acc | **0.2376** | ±0.0037 |
36
+ | - Humanities | acc | 0.2472 | ±0.0067 |
37
+ | - STEM | acc | 0.2245 | ±0.0074 |
38
+ | - Social Sciences | acc | 0.2327 | ±0.0076 |
39
+ | - Other | acc | 0.2430 | ±0.0077 |
40
+ | **GSM8K** | exact_match | **0.0240** | ±0.0048 |
41
+ | **HellaSwag** | acc_norm | **0.4430** | ±0.0157 |
42
+ | **ARC-Easy** | acc_norm | **0.5450** | ±0.0158 |
43
+ | **PIQA** | acc_norm | **0.6770** | ±0.0148 |
44
+ | **WinoGrande** | acc | **0.5210** | ±0.0158 |
45
+
46
  ### Key Innovation: The Asterisk Operator (★-operator)
47
 
48
  The **Asterisk Operator** performs local parallel state evolution through point-wise transformations:
 
170
  """
171
  ```
172
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
173
  **Note**: These are preliminary results with sample limits. Full evaluation pending.
174
 
175
  ## Quick Start
 
286
  author={NoesisLab},
287
  year={2026},
288
  publisher={Huggingface},
289
+ url={https://huggingface.co/NoesisLab/Asterisk}
290
  }
291
  ```
292