Update README.md

README.md

@@ -34,11 +34,6 @@ multiple-choice questions** drawn from TCM licensing-exam material, plus a
 **5,250-question subset answered by 102 licensed TCM practitioners**, giving a human
 reference point for the same items a model is scored on.
 
-The human subset is what distinguishes this from a plain exam dump. Every question in it
-carries the individual answers of the doctors who attempted it, their perceived
-difficulty rating, and their categorization of the question — so model accuracy can be
-compared against a measured human baseline rather than an assumed passing score.
-
 ## Configs
 
 ### `qa` (default) — 33,872 rows
@@ -54,7 +49,7 @@ compared against a measured human baseline rather than an assumed passing score.
 | `is_multi_answer` | bool | True for 2,308 rows (6.8%). |
 | `explanation` | string / null | Reference rationale where the source provided one (20,692 rows). |
 | `topic` | string / null | Free-text topic label (23,779 rows, 5,738 distinct). Uncontrolled vocabulary — see Limitations. |
-| `difficulty_raw` | int64 | Source-provided integer. **Not a calibrated difficulty signal** — see Limitations. |
+| `difficulty_raw` | int64 | Source-provided integer. |
 
 ### `doctor_annotated` — 5,250 rows, 15,151 human answers
 
@@ -109,74 +104,9 @@ latter is the appropriate baseline for most model-vs-human claims.
 Doctors answered under exam-like conditions without reference material. Majority vote on
 a 3-rater panel uses `num_correct_raters >= 2`.
 
-## Construction
-
-Questions were aggregated from Chinese TCM licensing-exam preparation material and from
-publicly released TCM exam datasets, then normalized to a uniform A–E multiple-choice
-schema. The following cleaning was applied to produce this release:
-
-| Step | Rows affected |
-|---|---|
-| Input | 38,279 |
-| Dropped — stem was a leaked answer key or explanation fragment, not a question | 14 |
-| Dropped — a correct option's text was duplicated verbatim by a distractor (unanswerable) | 34 |
-| Dropped — identical stem and options appearing with contradictory answer keys (32 groups) | 77 |
-| Dropped — exact duplicates of an earlier row (same stem, options and answer) | 4,282 |
-| **Output** | **33,872** |
-
-Text normalization applied to every retained field: literal `\"` escape artifacts
-unescaped, HTML markup and entities stripped, newlines and runs of whitespace collapsed
-to single spaces, leading/trailing whitespace removed, multi-answer keys sorted into
-canonical order. Empty `explanation` and `topic` strings became nulls.
-
-The internal source-document path recorded per question in the working collection was
-removed before release; it contained collection-workflow file paths and the names of
-individual contributors.
-
-## Provenance and attribution
-
-TCMQA was assembled, cleaned and released by [TechTCM](https://techtcm.com/), including
-the collection of the human practitioner annotations.
-
-Roughly 29% of the pre-cleaning rows were derived from publicly released datasets, and
-the remainder from an exam-preparation bank collected by TechTCM. Because the per-row
-source path was removed, upstream attribution is given here at the dataset level:
-
-- [`FreedomIntelligence/TCM-Text-Exams`](https://huggingface.co/datasets/FreedomIntelligence/TCM-Text-Exams)
-- `FreedomIntelligence/2023_Pharmacist_Licensure_Examination-TCM_track`
-- [`JTBTechnology/tcm_exam_questions`](https://huggingface.co/datasets/JTBTechnology/tcm_exam_questions)
-
-If you redistribute or build on TCMQA, carry these attributions forward. Users with
-commercial intentions should check the terms of the upstream datasets directly rather
-than relying on this release's license alone.
-
-## Limitations
-
-- **`difficulty_raw` is not a usable difficulty label.** It is uniformly `5` for every
-  row imported from an external dataset and ranges 0–9 for internally collected rows, so
-  it mostly encodes provenance rather than difficulty. It is retained only for
-  traceability. For a real difficulty signal, use `perceived_difficulty` or
-  `num_correct_raters` in the `doctor_annotated` config.
-- **`topic` is an uncontrolled vocabulary.** 5,738 distinct free-text strings at wildly
-  varying granularity (from `中医基础理论` down to `胃痛`), and absent on 10,093 rows.
-  Treat it as a weak hint, not a taxonomy. The `categories` field in
-  `doctor_annotated` is a proper 10-way scheme, but it reflects each rater's judgment and
-  raters sometimes disagree.
-- **Answer keys are as-published by the source material** and have not been independently
-  re-adjudicated by a clinician. The 32 contradictory groups removed above are the cases
-  where the collection disagreed with itself; single-source errors would not be caught.
-- **Mixed orthography.** Rows sourced from Taiwanese exams use traditional characters;
-  most rows use simplified. No conversion was applied.
-- **218 rows in `doctor_annotated` share a `qa_id` with another row** — the same question
-  was administered to more than one panel. Deduplicate on `qa_id` if you need one row per
-  distinct question.
-- **Not medical advice.** These are exam items. Performance here says something about
-  recall of TCM exam content and nothing about clinical safety or competence.
-
 ## Credit
 
-Built and maintained by **[TechTCM](https://techtcm.com/)**. If you use TCMQA in your
-work, please cite it and link back to <https://techtcm.com/>.
+Built and maintained by **[TechTCM](https://techtcm.com/)**.
 
 ## License