Add task_categories metadata

This PR adds the missing `task_categories` field to the dataset card YAML metadata. Since VoxParity is a benchmark for evaluating voice agents on paralinguistic and acoustic cues rather than a standard supervised task, `other` is the most appropriate category. No other changes are made to the README content.

README.md

@@ -1,8 +1,10 @@
 ---
-license: cc-by-4.0
-pretty_name: VoxParity development split
 language:
 - en
+license: cc-by-4.0
+size_categories:
+- n<1K
+pretty_name: VoxParity development split
 tags:
 - audio
 - speech
@@ -11,8 +13,8 @@ tags:
 - voice-agents
 - tool-calling
 - paralinguistics
-size_categories:
-- n<1K
+task_categories:
+- other
 ---
 
 > Hugging Face copy of the VoxParity development split v1.0.1 (paper: https://arxiv.org/abs/2609.35922; code and leaderboard: https://github.com/bhavik-mangla/voxparity-bench, https://bhavik-mangla.github.io/voxparity-bench/; archive: https://doi.org/10.5281/zenodo.23008159). Only this README gains the metadata header above; every other file is byte-identical to the Zenodo release.
@@ -159,4 +161,4 @@ text of the points below.
 Developing and checking voice agents before an organiser-run evaluation, and
 reproducing the paper's per-system numbers on the split. Not for training
 models, and not a substitute for the held-out evaluation: 40 items cannot carry
-a leaderboard claim on their own.
+a leaderboard claim on their own.