zackyalgiffari commited on
Commit
71bd3f9
·
verified ·
1 Parent(s): d5b5dfb

Upload nexus-decision-data

Browse files
Files changed (11) hide show
  1. .gitattributes +1 -0
  2. README.md +52 -0
  3. calib.jsonl +0 -0
  4. labels_fast.jsonl +0 -0
  5. labels_think.jsonl +0 -0
  6. public.jsonl +0 -0
  7. scenarios.jsonl +0 -0
  8. stats.json +9 -0
  9. test.jsonl +0 -0
  10. train.jsonl +3 -0
  11. val.jsonl +0 -0
.gitattributes CHANGED
@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ train.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ language: [en, id, es, de, fr, pt, ja]
4
+ task_categories: [text-classification]
5
+ tags: [decision-making, routing, triage, calibration, distillation, synthetic]
6
+ size_categories: [10K<n<100K]
7
+ configs:
8
+ - config_name: train
9
+ data_files: train.jsonl
10
+ - config_name: test
11
+ data_files: test.jsonl
12
+ ---
13
+
14
+ # nexus-decision-data
15
+
16
+ This is the training and evaluation data for
17
+ [zackyalgiffari/nexus-decision-4b](https://huggingface.co/zackyalgiffari/nexus-decision-4b), an open-weight
18
+ typed decision model. It holds typed decision questions (`choice`, `score`, `binary`) about an
19
+ application state, with soft labels from a teacher model.
20
+
21
+ | File | Contents |
22
+ |---|---|
23
+ | `scenarios.jsonl` | Synthetic scenarios: `state` plus 1-4 typed `questions`, in 12 domains. Written by DeepSeek-V4.1-Flash. |
24
+ | `public.jsonl` | BANKING77 and CLINC150 texts converted to typed questions, with `gold` labels |
25
+ | `labels_fast.jsonl` | Teacher soft distributions (top-20 logprobs, averaged over 3 option orders) |
26
+ | `labels_think.jsonl` | Teacher reasoning traces (thinking mode) with the displayed option order and the answer |
27
+ | `train.jsonl` / `val.jsonl` | Rendered training records (`user_text`, `labels`, `target`, optional `reasoning`) |
28
+ | `calib.jsonl` / `test.jsonl` | Evaluation items (`state`, `q`, `labels`, `teacher`, `gold`) |
29
+ | `stats.json` | Split statistics |
30
+
31
+ | Statistic | Value |
32
+ |---|---|
33
+ | calib_items | 804 |
34
+ | test_items | 1,786 |
35
+ | think_dropped_disagree | 454 |
36
+ | think_dropped_long | 15 |
37
+ | train_fast | 12,336 |
38
+ | train_think | 4,005 |
39
+ | val_fast | 804 |
40
+
41
+ ## Provenance and licensing
42
+
43
+ - **Synthetic scenarios and labels.** They were generated with the DeepSeek API (`deepseek-flash`)
44
+ and are released under CC-BY-4.0. They are model-generated: they may contain mistakes, and any
45
+ names and IDs in them are fictional.
46
+ - **BANKING77-derived items** (`source: banking77`). The texts come from
47
+ [PolyAI/banking77](https://huggingface.co/datasets/PolyAI/banking77), CC-BY-4.0.
48
+ - **CLINC150-derived items** (`source: clinc150`). The texts come from
49
+ [clinc/clinc_oos](https://huggingface.co/datasets/clinc/clinc_oos), CC-BY-3.0 (Larson et al., 2019).
50
+
51
+ The test split of the public data is the original datasets' test split. It was never used for
52
+ training.
calib.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
labels_fast.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
labels_think.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
public.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
scenarios.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
stats.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "calib_items": 804,
3
+ "test_items": 1786,
4
+ "think_dropped_disagree": 454,
5
+ "think_dropped_long": 15,
6
+ "train_fast": 12336,
7
+ "train_think": 4005,
8
+ "val_fast": 804
9
+ }
test.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
train.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50aded9f79f68445518a041b6d11e13cd8ecdb85e90a375ea44b10f52bf2ecca
3
+ size 27517671
val.jsonl ADDED
The diff for this file is too large to render. See raw diff