Irfanuruchi commited on
Commit
4af80be
·
verified ·
1 Parent(s): c46728e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +63 -5
README.md CHANGED
@@ -11,12 +11,70 @@ tags:
11
  - reasoning
12
  ---
13
 
14
- # BuildEng Dataset
15
 
16
- This is the main dataset holder for the BuildEng 32B branch.
17
 
18
- The 32B branch is planned as the flagship large-scale civil/building engineering dataset for deeper structural reasoning, longer workflows, construction-stage analysis, and broader engineering coverage.
19
 
20
- Files will be added as the 32B branch is finalized.
21
 
22
- Author: Irfan Uruchi
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  - reasoning
12
  ---
13
 
14
+ # BuildEng V8 145K
15
 
16
+ BuildEng V8 145K is a civil and structural engineering reasoning dataset built for fine-tuning conservative engineering assistant models.
17
 
18
+ The dataset contains 145,117 validated chat-format samples focused on structural judgment, inspection-first reasoning, safe uncertainty handling, and practical building-engineering logic.
19
 
20
+ This dataset is not designed to make a model give final structural approvals. It is designed to teach an assistant to reason carefully, separate field symptoms from root causes, avoid unsafe certainty, request missing information, and recommend qualified engineering review when the situation requires it.
21
 
22
+ ## Dataset Focus
23
+
24
+ The dataset covers reinforced concrete beams, steel beams, slabs, columns, foundations, retaining walls, masonry walls, roof systems, temporary shoring, construction sequencing, settlement, cracking, corrosion, deflection, load-path changes, repair decisions, contradictory inspection reports, and failure progression.
25
+
26
+ It also includes supporting building-engineering topics such as HVAC airflow, ducts, hydronics, electrical load checks, and safety-related MEP reasoning.
27
+
28
+ ## Reasoning Behavior
29
+
30
+ The dataset trains models to follow this pattern:
31
+
32
+ Observation → preliminary reasoning → missing information → risk check → conservative next action.
33
+
34
+ Key behaviors include:
35
+
36
+ * inspection-first decision making
37
+ * symptom vs cause separation
38
+ * uncertainty-aware answers
39
+ * refusal of unsafe conclusions
40
+ * contradiction handling
41
+ * repair decision logic
42
+ * load-path awareness
43
+ * civil formula reasoning
44
+ * integrated calculation plus safety limits
45
+ * strength vs serviceability separation
46
+
47
+ ## Format
48
+
49
+ Each sample uses chat messages format:
50
+
51
+ ```json
52
+ {
53
+ "messages": [
54
+ {
55
+ "role": "system",
56
+ "content": "You are BuildEng, a conservative civil and structural engineering assistant."
57
+ },
58
+ {
59
+ "role": "user",
60
+ "content": "..."
61
+ },
62
+ {
63
+ "role": "assistant",
64
+ "content": "..."
65
+ }
66
+ ]
67
+ }
68
+ ```
69
+
70
+ ## Intended Use
71
+
72
+ This dataset is intended for supervised fine-tuning of civil/structural engineering assistant models, especially models that should behave cautiously when dealing with incomplete field information or potentially unsafe structural conditions.
73
+
74
+ It was used for BuildEng V8 model development and is intended for further 3B and 32B QLoRA training.
75
+
76
+ ## Safety Note
77
+
78
+ The dataset is educational and assistant-training material. Models trained on it should not replace licensed structural engineers, site inspections, building-code checks, or professional engineering judgment.
79
+
80
+ Final structural decisions must be made by qualified professionals using project-specific drawings, calculations, inspections, standards, and local regulations.