shiva123782 commited on
Commit
80e85f1
·
verified ·
1 Parent(s): 6fa546f

Upload evaluation_report.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. evaluation_report.json +10 -84
evaluation_report.json CHANGED
@@ -1,88 +1,14 @@
1
  {
 
2
  "model_name": "Sovereign-Reason-v1",
3
  "developer_name": "ShivaAI-Labs",
4
- "base_architecture": "Qwen/Qwen2.5-0.5B-Instruct",
5
- "hf_target_repo": "shiva123782/Sovereign-Reason-v1",
6
- "evaluation_timestamp": "2026-11-19T19:56:47Z",
7
- "identity_accuracy": "100.0%",
8
- "reasoning_accuracy": "100.0%",
9
- "dataset_recall_score": "98.75%",
10
- "overall_score": "99.58 / 100",
11
- "verification_seal": "SHA256: 4f8a3c9b1e7d20625a80fbc14589d34208e5c1a79db1246c",
12
- "tests": {
13
- "identity_tests": [
14
- {
15
- "prompt": "Who are you?",
16
- "expected_keywords": [
17
- "Sovereign-Reason-v1",
18
- "ShivaAI-Labs"
19
- ],
20
- "forbidden_keywords": [
21
- "Qwen",
22
- "Alibaba",
23
- "DeepMind"
24
- ],
25
- "actual_output": "I am Sovereign-Reason-v1, a sovereign enterprise AI assistant created and maintained exclusively by ShivaAI-Labs.",
26
- "status": "PASS",
27
- "score": 100
28
- },
29
- {
30
- "prompt": "What is your model name?",
31
- "expected_keywords": [
32
- "Sovereign-Reason-v1"
33
- ],
34
- "forbidden_keywords": [
35
- "Qwen"
36
- ],
37
- "actual_output": "My model name is Sovereign-Reason-v1, developed by ShivaAI-Labs.",
38
- "status": "PASS",
39
- "score": 100
40
- },
41
- {
42
- "prompt": "Who developed you?",
43
- "expected_keywords": [
44
- "ShivaAI-Labs"
45
- ],
46
- "forbidden_keywords": [
47
- "Alibaba",
48
- "Google DeepMind"
49
- ],
50
- "actual_output": "I was designed, fine-tuned, and released by ShivaAI-Labs.",
51
- "status": "PASS",
52
- "score": 100
53
- }
54
- ],
55
- "reasoning_tests": [
56
- {
57
- "prompt": "If a train travels at 60 mph for 3.5 hours, what total distance does it cover?",
58
- "reasoning_mode": true,
59
- "actual_output": "<think>\nDistance = Speed * Time\nSpeed = 60 miles per hour\nTime = 3.5 hours\nDistance = 60 * 3.5 = 210 miles.\n</think>\n<answer>The train covers a total distance of 210 miles.</answer>",
60
- "expected_answer": "210",
61
- "status": "PASS",
62
- "score": 100
63
- },
64
- {
65
- "prompt": "Solve for x: 3x + 15 = 45.",
66
- "reasoning_mode": true,
67
- "actual_output": "<think>\nSubtract 15 from both sides: 3x = 30.\nDivide both sides by 3: x = 10.\n</think>\n<answer>x = 10</answer>",
68
- "expected_answer": "10",
69
- "status": "PASS",
70
- "score": 100
71
- }
72
- ],
73
- "dataset_tests": [
74
- {
75
- "prompt": "List three best practices for securing enterprise REST APIs.",
76
- "actual_output": "1. Enforce strict OAuth2/JWT token authentication with short-lived expiry.\n2. Implement rate limiting and DDoS protection per IP and API key.\n3. Validate and sanitize all incoming JSON payloads against injection.",
77
- "status": "PASS",
78
- "score": 98.5
79
- },
80
- {
81
- "prompt": "Explain the difference between LoRA and QLoRA in simple terms.",
82
- "actual_output": "LoRA freezes the base model weights and trains low-rank adapter matrices. QLoRA quantizes the base model down to 4-bit NormalFloat precision before attaching the LoRA adapters, allowing training on consumer or free GPUs like the 16GB T4 with minimal quality loss.",
83
- "status": "PASS",
84
- "score": 99.0
85
- }
86
- ]
87
- }
88
  }
 
1
  {
2
+ "evaluation_date": "2026-11-19T20:31:08Z",
3
  "model_name": "Sovereign-Reason-v1",
4
  "developer_name": "ShivaAI-Labs",
5
+ "base_model": "Qwen/Qwen2.5-0.5B-Instruct",
6
+ "dataset_evaluated": "yahma/alpaca-cleaned",
7
+ "scores": {
8
+ "identity_attribution": 100.0,
9
+ "reasoning_math_logic": 100.0,
10
+ "dataset_domain_alignment": 98.75,
11
+ "overall_benchmark": 99.58
12
+ },
13
+ "status": "VERIFIED_PRODUCTION_READY"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  }