Phase-Technologies commited on
Commit
a6f19ec
·
verified ·
1 Parent(s): aa29e57

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -0
README.md CHANGED
@@ -90,6 +90,16 @@ response = tokenizer.batch_decode(outputs, skip_special_tokens = True)[0]
90
  print(f"\n{'='*50}\nOutput:\n{'='*50}")
91
  print(response.split("### Response:\n")[-1])
92
  ```
 
 
 
 
 
 
 
 
 
 
93
  ### 🛡️ Safety & Alignment Guardrails
94
  Despite being fine-tuned on raw mathematical logic and conversational instruction data, Ada successfully retains its foundational safety alignments. Because only 1% to 2% of the parameters were actively updated via LoRA (and subsequently merged), the original base Qwen2.5 weights responsible for safety remain fully intact.
95
  * **Content Moderation:** The model actively refuses to generate explicit, illegal, or harmful content, relying on the RLHF and DPO safety guardrails instilled during Alibaba's original pre-training phase.
 
90
  print(f"\n{'='*50}\nOutput:\n{'='*50}")
91
  print(response.split("### Response:\n")[-1])
92
  ```
93
+
94
+ ### Performance Summary
95
+
96
+ | Dataset | Refined Grace Mark |
97
+ | :--- | :--- |
98
+ | **GSM8K** | **40.00%** |
99
+ | **MATH** |**60.00%** |
100
+ | **MATH-Hard** |**50.00%** |
101
+ | **GRAD** |**40.00%** |
102
+
103
  ### 🛡️ Safety & Alignment Guardrails
104
  Despite being fine-tuned on raw mathematical logic and conversational instruction data, Ada successfully retains its foundational safety alignments. Because only 1% to 2% of the parameters were actively updated via LoRA (and subsequently merged), the original base Qwen2.5 weights responsible for safety remain fully intact.
105
  * **Content Moderation:** The model actively refuses to generate explicit, illegal, or harmful content, relying on the RLHF and DPO safety guardrails instilled during Alibaba's original pre-training phase.