Upload README.md with huggingface_hub

README.md

@@ -9,60 +9,30 @@ tags:
 - human-difficulty
 - orthographic-constraints
 - llm-evaluation
-- cognitive-benchmark
 size_categories:
 - n<1K
 task_categories:
 - text-generation
 pretty_name: NYT Spelling Bee Human Difficulty
-dataset_info:
-  features:
-  - name: date
-    dtype: string
-  - name: center_letter
-    dtype: string
-  - name: outer_letters
-    list: string
-  - name: answer_words
-    list: string
-  - name: answer_user_counts
-    list: int32
-  splits:
-  - name: train
-    num_bytes: 40186
-    num_examples: 58
-  download_size: 42419
-  dataset_size: 40186
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
 ---
 
 # NYT Spelling Bee — Human Difficulty Dataset
 
 Human solve-frequency data for 58 New York Times Spelling Bee puzzles (June–July 2025), sampled from 10,000 users per puzzle. Serves as ground truth for evaluating LLM orthographic constraint satisfaction.
 
-## Dataset Description
+## Task
 
-Each row is one daily puzzle. The Spelling Bee task requires generating valid English words using only 7 given letters, with one designated center letter that must appear in every word.
-
-The `answer_user_counts` field records how many of 10,000 sampled users found each answer word, providing a continuous human difficulty signal per word.
+The Spelling Bee requires generating valid English words using only 7 given letters. One designated center letter must appear in every word.
 
 ## Schema
 
 | Column | Type | Description |
 |--------|------|-------------|
 | `date` | string | Puzzle date (YYYY-MM-DD) |
-| `puzzle_id` | int | NYT puzzle ID |
 | `center_letter` | string | Required center letter |
 | `outer_letters` | list[str] | 6 outer letters |
-| `num_answers` | int | Number of valid answer words |
 | `answer_words` | list[str] | All valid answer words |
 | `answer_user_counts` | list[int] | Users (of 10,000) who found each word |
-| `sample_size` | int | Sample size (10,000) |
-| `total_players` | int | Total players that day |
 
 ## Usage
 
@@ -74,21 +44,12 @@ puzzle = ds["train"][0]
 
 # Human solve rate per word
 for word, count in zip(puzzle["answer_words"], puzzle["answer_user_counts"]):
-    print(f"{word}: {count / puzzle['sample_size']:.1%}")
+    print(f"{word}: {count / 10_000:.1%}")
 ```
 
-## Key Statistics
-
-- **58 puzzles** (2025-06-02 to 2025-07-29)
-- **20–70+ answer words per puzzle**
-- **10,000 sampled users** per puzzle
-- **50,000–100,000+ total players** per puzzle
-
 ## Associated Paper
 
-This dataset supports the following paper:
-
-> Bryan Tuck and Rakesh Verma. "Evaluating LLM Orthographic Constraint Satisfaction with the NYT Spelling Bee." In *Proceedings of LREC-COLING 2026*.
+> Bryan E. Tuck and Rakesh M. Verma. "Orthographic Constraint Satisfaction and Human Difficulty Alignment in Large Language Models." arXiv:2511.21086, 2025.
 
 ## License
 
@@ -97,10 +58,13 @@ The puzzle structure and answer data are derived from the New York Times Spellin
 ## Citation
 
 ```bibtex
-@inproceedings{tuck2026spelling,
-  title={Evaluating {LLM} Orthographic Constraint Satisfaction with the {NYT} Spelling Bee},
-  author={Tuck, Bryan and Verma, Rakesh},
-  booktitle={Proceedings of the International Conference on Language Resources and Evaluation (LREC-COLING)},
-  year={2026}
+@misc{tuck2025orthographicconstraintsatisfactionhuman,
+      title={Orthographic Constraint Satisfaction and Human Difficulty Alignment in Large Language Models},
+      author={Bryan E. Tuck and Rakesh M. Verma},
+      year={2025},
+      eprint={2511.21086},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2511.21086},
 }
 ```