sigdelakshey commited on
Commit
561075f
·
verified ·
1 Parent(s): 749b72b

Clarify expected nulls in fault viewer fields

Browse files
Files changed (1) hide show
  1. README.md +10 -1
README.md CHANGED
@@ -37,7 +37,7 @@ All tasks are synthetic and generated with deterministic seeds to ensure reprodu
37
  ## Repository and Evaluator
38
 
39
  - Project repository (code + evaluator + baselines):
40
- https://github.com/akgitrepos/toolmisusebench
41
  - Recommended evaluation flow uses the project CLI and harness.
42
 
43
  ## Supported Evaluation Use Cases
@@ -99,6 +99,15 @@ Supported fault types:
99
 
100
  Faults are declaratively specified per task and replayed deterministically.
101
 
 
 
 
 
 
 
 
 
 
102
  ## Data Generation
103
 
104
  Generated synthetically using deterministic templates, seeded randomization, and task-level coherence checks.
 
37
  ## Repository and Evaluator
38
 
39
  - Project repository (code + evaluator + baselines):
40
+ `https://github.com/akgitrepos/toolmisusebench`
41
  - Recommended evaluation flow uses the project CLI and harness.
42
 
43
  ## Supported Evaluation Use Cases
 
99
 
100
  Faults are declaratively specified per task and replayed deterministically.
101
 
102
+ ## Viewer Note on Null Values
103
+
104
+ In the Hugging Face table viewer, nested fields inside `fault_plan.trigger` and `fault_plan.payload`
105
+ may appear as `null` for some rows.
106
+
107
+ This is expected: different fault types use different subsets of fields, and the viewer displays a
108
+ unified schema across all rows. A `null` value in this context typically means "not applicable for
109
+ this fault type," not missing or corrupted data.
110
+
111
  ## Data Generation
112
 
113
  Generated synthetically using deterministic templates, seeded randomization, and task-level coherence checks.