filippogreppi commited on
Commit
947f4d8
·
verified ·
1 Parent(s): 34ea100

Upload the training files of the second-phase students

Browse files
.gitattributes CHANGED
@@ -58,3 +58,7 @@ 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
+ pilener_tutor_free_cot.jsonl filter=lfs diff=lfs merge=lfs -text
62
+ pilener_tutor_free_cot_medgemma.jsonl filter=lfs diff=lfs merge=lfs -text
63
+ pilener_tutor_structured_cot.jsonl filter=lfs diff=lfs merge=lfs -text
64
+ pilener_tutor_structured_cot_medgemma.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags: [named-entity-recognition, chain-of-thought, sft]
3
+ language: [en]
4
+ pretty_name: UniversalMedNER – Training Files of the Free vs Structured Reasoning Students
5
+ ---
6
+
7
+ # UniversalMedNER — Training Files of the Free vs Structured Reasoning Students
8
+
9
+ Part of **UniversalMedNER**, the code, data and models of the bachelor thesis *Structuring Universal Chain-of-thought for Named Entity Recognition* (Department of Computer Science and Engineering, University of Bologna, 2026).
10
+
11
+ The supervised fine-tuning files of the second-phase students (the `*-cot-lora` adapters of the organization), built
12
+ from the train split of `disi-unibo-nlp-students/pilener-teacher-cot-agreement-v1` (3,822 queries). Every file has the
13
+ same queries, in the same order, and the same target answer (the teacher's `answer`); only the reasoning differs.
14
+
15
+ Each row: `id` (the query id of the source dataset) and `messages`, a user turn (the prompt) and an assistant turn
16
+ `<think>\n{reasoning}</think>\n\n{answer}`, where the answer is a JSON object mapping every requested type to its
17
+ mentions. No system prompt; the loss is computed on the assistant turn only.
18
+
19
+ | File | Reasoning | Prompt | Built with |
20
+ |---|---|---|---|
21
+ | `pilener_tutor_free_cot.jsonl` | the teacher's free reasoning (`free_cot`) | `free_prompt` | `05_openvocab_sft/build_tutor_cot_dataset.py` |
22
+ | `pilener_tutor_structured_cot.jsonl` | the teacher's structured reasoning (`structured_cot`): `TYPES`, `MENTIONS`, `EDGE` | `structured_prompt` | `05_openvocab_sft/build_tutor_cot_dataset.py` |
23
+ | `pilener_tutor_noedge_cot.jsonl` | the structured reasoning without the `EDGE` line | `structured_prompt` without the `EDGE` instruction | `05_openvocab_sft/make_noedge_data.py` |
24
+ | `pilener_tutor_noedge_inv_cot.jsonl` | the structured reasoning without `EDGE`, with `MENTIONS` before `TYPES` | the same edit, with the two instructions swapped | `05_openvocab_sft/make_noedge_data.py --invert` |
25
+ | `pilener_tutor_noreason_cot.jsonl` | none (empty thinking block) | `free_prompt` | `05_openvocab_sft/make_noreason_data.py` |
26
+ | `pilener_tutor_free_cot_medgemma.jsonl` | as the free file, in MedGemma's thought block: `<unused94>thought\n{reasoning}<unused95>{answer}` | `free_prompt` | `05_openvocab_sft/make_medgemma_tutor_data.py` |
27
+ | `pilener_tutor_structured_cot_medgemma.jsonl` | as the structured file, in MedGemma's thought block | `structured_prompt` | `05_openvocab_sft/make_medgemma_tutor_data.py` |
28
+
29
+ The variants without `EDGE` are derived from the existing structured traces (the line is removed), not regenerated
30
+ by the teacher; they are additional experiments, as is the file without reasoning.
pilener_tutor_free_cot.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aae82942130ba553070d2b3c038a3325a03c87aebb8a8891e205b8287b57d336
3
+ size 13625476
pilener_tutor_free_cot_medgemma.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45a170356109f8a13e36cebc56cc5e04d0e172aaf5e6b94de40b188663bc2537
3
+ size 13648408
pilener_tutor_noedge_cot.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
pilener_tutor_noedge_inv_cot.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
pilener_tutor_noreason_cot.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
pilener_tutor_structured_cot.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:577100a88ebb058edff12de01fd36ce19f241d0b9feb605fa2f74fc2442200a8
3
+ size 11502230
pilener_tutor_structured_cot_medgemma.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f06d6ddb96b9f612c0607ea5a7c398927995df32e1c6b5267d5ea4ac85c57789
3
+ size 11525162