ForumCore commited on
Commit
ea7f2ab
·
verified ·
1 Parent(s): 300cde4

ops: DotCheck Apache-2.0 open heads + card

Browse files
Files changed (2) hide show
  1. README.md +10 -10
  2. text_stack_head_v11.npz +3 -0
README.md CHANGED
@@ -24,7 +24,7 @@ base_model:
24
  - hfl/chinese-macbert-base
25
  base_model_relation: adapter
26
  model-index:
27
- - name: Valla (inhouse-text@9)
28
  results:
29
  - task:
30
  type: text-classification
@@ -36,13 +36,13 @@ model-index:
36
  metrics:
37
  - name: mean_P_AI_human
38
  type: mean_score_human
39
- value: 0.045
40
  - name: mean_P_AI_ai
41
  type: mean_score_ai
42
- value: 0.909
43
  - name: balanced_accuracy
44
  type: balanced_accuracy
45
- value: 0.980
46
  source:
47
  name: text_gates_v9 / Data.json
48
  url: https://dotcheck.ai/docs
@@ -55,9 +55,9 @@ Apache-2.0 text AI-likeness heads for DotCheck (English, Chinese, and six Latin
55
  | Field | Value |
56
  |-------|--------|
57
  | Hub id | `DotCheck/valla-text-v9` |
58
- | Wire id (EN) | `inhouse-text@9` |
59
  | Label | Valla |
60
- | EN artifact | `text_stack_head_v9.npz` |
61
  | Latin feature bases | [`Oxidane/tmr-ai-text-detector`](https://huggingface.co/Oxidane/tmr-ai-text-detector) (MIT), [`fakespot-ai/roberta-base-ai-text-detection-v1`](https://huggingface.co/fakespot-ai/roberta-base-ai-text-detection-v1) (Apache-2.0) |
62
  | Chinese feature base | [`hfl/chinese-macbert-base`](https://huggingface.co/hfl/chinese-macbert-base) (Apache-2.0; `cls_emb`) |
63
  | Heads | EN + `zh_v1` / `es_v2` / `pt_BR_v1` / `fr_v2` / `it_v1` / `de_v1` / `nl_v1` (separate `.npz`) |
@@ -122,9 +122,9 @@ No published holdout from live ChatGPT/Claude scrapes. Evidence: `text_gates_v9.
122
 
123
  | Metric | Target | Measured |
124
  |--------|--------|---------:|
125
- | mean P(AI) \| human | ≤ 0.12 | **0.045** |
126
- | mean P(AI) \| AI | ≥ 0.85 | **0.909** |
127
- | bal_acc @ thr | ≥ 0.90 | **0.980** |
128
 
129
  Also: beat prior `@6` holdout; wiki-monitor mean ~\<0.001; OOD / RAID-lite protocol OK (`TEXT_GATES_OK`).
130
 
@@ -166,4 +166,4 @@ Also: beat prior `@6` holdout; wiki-monitor mean ~\<0.001; OOD / RAID-lite proto
166
 
167
  ## Citation
168
 
169
- [`CITATION.cff`](CITATION.cff) · wire `inhouse-text@9` / Valla@9 · https://dotcheck.ai/docs
 
24
  - hfl/chinese-macbert-base
25
  base_model_relation: adapter
26
  model-index:
27
+ - name: Valla (inhouse-text@11)
28
  results:
29
  - task:
30
  type: text-classification
 
36
  metrics:
37
  - name: mean_P_AI_human
38
  type: mean_score_human
39
+ value: 0.005
40
  - name: mean_P_AI_ai
41
  type: mean_score_ai
42
+ value: 0.973
43
  - name: balanced_accuracy
44
  type: balanced_accuracy
45
+ value: 0.986
46
  source:
47
  name: text_gates_v9 / Data.json
48
  url: https://dotcheck.ai/docs
 
55
  | Field | Value |
56
  |-------|--------|
57
  | Hub id | `DotCheck/valla-text-v9` |
58
+ | Wire id (EN) | `inhouse-text@11` |
59
  | Label | Valla |
60
+ | EN artifact | `text_stack_head_v11.npz` |
61
  | Latin feature bases | [`Oxidane/tmr-ai-text-detector`](https://huggingface.co/Oxidane/tmr-ai-text-detector) (MIT), [`fakespot-ai/roberta-base-ai-text-detection-v1`](https://huggingface.co/fakespot-ai/roberta-base-ai-text-detection-v1) (Apache-2.0) |
62
  | Chinese feature base | [`hfl/chinese-macbert-base`](https://huggingface.co/hfl/chinese-macbert-base) (Apache-2.0; `cls_emb`) |
63
  | Heads | EN + `zh_v1` / `es_v2` / `pt_BR_v1` / `fr_v2` / `it_v1` / `de_v1` / `nl_v1` (separate `.npz`) |
 
122
 
123
  | Metric | Target | Measured |
124
  |--------|--------|---------:|
125
+ | mean P(AI) \| human | ≤ 0.12 | **0.005** |
126
+ | mean P(AI) \| AI | ≥ 0.85 | **0.973** |
127
+ | bal_acc @ thr | ≥ 0.90 | **0.986** |
128
 
129
  Also: beat prior `@6` holdout; wiki-monitor mean ~\<0.001; OOD / RAID-lite protocol OK (`TEXT_GATES_OK`).
130
 
 
166
 
167
  ## Citation
168
 
169
+ [`CITATION.cff`](CITATION.cff) · wire `inhouse-text@11` / Valla@9 · https://dotcheck.ai/docs
text_stack_head_v11.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:941f8106cdcc9f10acd2ca587e6f167cb3b02ce23dc3e72c8d350626f313de97
3
+ size 9074