Card: arXiv/GitHub links, fix load_dataset repo id, de-anonymize

README.md

@@ -164,6 +164,12 @@ configs:
 
 # PitchBench
 
+[![arXiv](https://img.shields.io/badge/arXiv-2605.26176-b31b1b.svg)](https://arxiv.org/abs/2605.26176)
+[![GitHub](https://img.shields.io/badge/GitHub-vaclisinc%2FPitchBench-181717.svg?logo=github)](https://github.com/vaclisinc/PitchBench)
+[![PyPI](https://img.shields.io/pypi/v/pitchbench.svg?logo=pypi&logoColor=white&label=PyPI)](https://pypi.org/project/pitchbench/)
+
+Paper: [PitchBench: Measuring Pitch Hearing in Audio-Language Models](https://arxiv.org/abs/2605.26176)
+
 A benchmark for testing what audio / acoustic signals **Audio Language Models (ALMs)** do
 and don't understand. PitchBench probes pitch perception across 29 controlled
 experiments — single-pitch ID, onsets/offsets, chords, sequences, contour, audio
@@ -179,18 +185,18 @@ question (`prompt*`) asked of the model, and the ground-truth answer fields
 from datasets import load_dataset
 
 # Load one experiment (configurations match experiment IDs)
-ds = load_dataset("pitchbench-authors/PitchBench", "pitchbench_a1_single_pitch_id", split="test")
+ds = load_dataset("vaclis/PitchBench", "pitchbench_a1_single_pitch_id", split="test")
 print(ds[0]["audio"], ds[0]["prompt_midi"], ds[0]["gt_midi"])
 
 # d7b has two audio columns: audio_1 (reference tone) and audio_2 (target tone)
-ds_d7b = load_dataset("pitchbench-authors/PitchBench",
+ds_d7b = load_dataset("vaclis/PitchBench",
                       "pitchbench_d7b_pitch_with_reference_split", split="test")
 print(ds_d7b[0]["audio_1"], ds_d7b[0]["audio_2"], ds_d7b[0]["gt_midi"])
 
 # Iterate every experiment
 import datasets
-for cfg in datasets.get_dataset_config_names("pitchbench-authors/PitchBench"):
-    ds = load_dataset("pitchbench-authors/PitchBench", cfg, split="test")
+for cfg in datasets.get_dataset_config_names("vaclis/PitchBench"):
+    ds = load_dataset("vaclis/PitchBench", cfg, split="test")
     print(cfg, len(ds))
 ```
 
@@ -407,8 +413,9 @@ Stimuli are generated deterministically from the configuration in
 `pitchbench.config` (`EVAL=True` benchmark constants). The subset published
 here is the seeded stratified sample used in the paper — reproduced by
 `apply_default_sampling(EXP_NAME, all_conds, None, seed=42)` over the output
-of each experiment's `build_conditions(...)`. Source code is provided as
-supplementary material with the submission (anonymized for review).
+of each experiment's `build_conditions(...)`. Source code is available at
+[github.com/vaclisinc/PitchBench](https://github.com/vaclisinc/PitchBench)
+(`pip install pitchbench`).
 
 ## License