Datasets:

chnln
/

seeing-is-not-sharing

license: cc-by-4.0
task_categories:
  - text-classification
language:
  - en
tags:
  - dialogue
  - common-ground
  - maptask
  - vision-language
size_categories:
  - 10K<n<100K

Seeing Is Not Sharing (SINS) Binary Common-Ground Judgment Dataset

Dataset Description

SINS is a 13,077-instance binary common-ground judgment (interpretation matching judgment) dataset, derived from the public Grounded Misunderstandings in MapTask (GMMT) annotations. Each instance asks whether the giver and follower interpret the referring expression as the same landmark. Rows preserve identifiers and context-window provenance while mapping GMMT status to gold_label: aligned becomes yes; pending and misunderstood become no.

Related Resources

Resource	Location
SINS code and prompts	GitHub
SINS paper	arXiv:2606.31719, to appear in SIGDIAL 2026
GMMT code	GitHub
GMMT dataset	Hugging Face
GMMT paper	arXiv:2511.03718, LREC 2026

Load with Datasets

from datasets import load_dataset

ds = load_dataset("chnln/seeing-is-not-sharing", split="train")

Data Fields

The release contains ref_id, dialogue_id, map_id, utt_id, transaction_id, context_transaction_ids, end_utt_id_of_context, timed_unit_ids, expression, status, and gold_label.

Excluded Material

The SINS HF dataset does not contain dialogue context. It does not contain MapTask maps, images, OCR, or image-derived text. It also does not contain generated prompts, model predictions, log-probabilities, or copies of the underlying MapTask source files. Users who download MapTask themselves may reconstruct context with the SINS code repository and a local GMMT checkout.

Licensing and Provenance

SINS is released under CC BY 4.0. It is derived from GMMT, which is also CC BY 4.0. Citation details for both papers are provided below. See NOTICE in the code repository for details.

Citation

SINS builds on the GMMT data and annotation scheme introduced in our LREC 2026 paper. If you use SINS, please cite both the SINS paper and the GMMT paper.

SINS: Task and Experiments

To appear in SIGDIAL 2026.

@misc{li2026seeing,
  title = {Seeing Is Not Sharing: Some Vision-Language Models Overestimate Common Ground in Asymmetric Dialogue},
  author = {Li, Nan and Gatt, Albert and Poesio, Massimo},
  year = {2026},
  eprint = {2606.31719},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL},
  url = {https://arxiv.org/abs/2606.31719},
  note = {To appear in SIGDIAL 2026}
}

GMMT: Data and Annotation Scheme

Published at LREC 2026.

@inproceedings{li2026grounded,
  title = {Grounded Misunderstandings in Asymmetric Dialogue: A Perspectivist Annotation Scheme for MapTask},
  author = {Li, Nan and Gatt, Albert and Poesio, Massimo},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
  month = {May},
  year = {2026},
  pages = {4988--5001},
  address = {Palma, Mallorca, Spain},
  publisher = {European Language Resources Association (ELRA)},
  editor = {Piperidis, Stelios and Bel, Núria and van den Heuvel, Henk and Ide, Nancy and Krek, Simon and Toral, Antonio},
  doi = {10.63317/59anbt78wyj7}
}