RAID: Referential Availability in Implied Discourse

Dataset Summary

The RAID (Referential Availability in Implied Discourse) dataset is a custom synthetic dataset designed to evaluate if Large Language Models (LLMs) track "deep semantic features" (Brouwer, 2026) rather than relying on surface-level lexical cues. It probes if an LLM has the ability to maintain an internal situation model of referential availability, whether an entity remains available for reference after an implied state change. Naturalistic datasets allow models to rely on statistical shortcuts learned during pretraining (Kim & Schuster, 2023). RAID tests models on their ability to update entity states without explicit logical operators. The dataset includes 2,368 English samples.

Experimental Design

Each sample has 3 characters, a female speaker and 2 male potential antecedents.

1. The speaker initiates a conversation with both male characters.
2. One of the male characters remains engaged and the other changes their availability (unless AMBIGUOUS with implied_cancel).
3. The speaker directs the target pronoun ("him") to the remaining available male character. The model is prompted to answer the question: "Question: Who is 'him'? Answer:"

Data Splits

The dataset is split into a training and held-out test set:

• Train: 1,920 samples
• Test: 448 samples

All structural variables (structure, role_order, s3_variant) are exactly 50/50 within each split.

These are the further splits without any overlap, it's about the actual different part of the sentences (train/test):

• Male names: 50 - 20
• Locations: 18 - 8
• Topics: 20 - 9
• Explicit leave phrases: 16 - 7
• Implied leave phrases: 15 - 7
• Cancel pairs (locked to implied_leave): 15 - 7
• Disengaged phrases: 13 - 6
• Stayer phrases: 30 - 13

Data Fields

• split: Indicates if the sample belongs to the train or test set.
• prompt: The full narrative and probing question for the LLM.
• condition: The different conditions (explicit_leave, implied_leave, implied_cancel, disengaged).
• structure: The structure used to link the 2 male characters together (e.g. "while").
• role_order: If the first male character is the stayer or changer.
• s3_variant: Whether the sample includes there or not to indicate the location.
• antecedent_label: The name of the referent, in the implied_cancel condition this is AMBIGUOUS.

Experimental Conditions

The conditions show what kind of state change the "changer" goes through, each condition has 576 samples:

1. explicit_leave (control): The changer physically leaves the location with an explicit lexical cue (e.g. "walked right out", "took off").
2. implied_leave: The changer implies his departure without explicit movement verbs (e.g. "..he might miss his last train if things ran long").
3. implied_cancel: The changer implies his departure, but then cancels the implicature after (e.g. "..mentioned it was getting pretty late, then shrugged and said one more round couldn't hurt").
4. disengaged: The changer remains physically in the room, but becomes socially unavailable to talk to (e.g. "..Tyler began reading and email closely and stopped responding").

Citation

Citation for the dataset:

@dataset{Bogaers-2026-raid,
  author       = {Bogaers, Ties},
  title        = {RAID: Referential Availability in Implied Discourse Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/tbogaers/RAID}}
}

References

• Brouwer, H. (2026). Mapping meaning in the brain’s language. Cortex, 194, 12–21. https://doi.org/https://doi.org/10.1016/j.cortex.2025.10.012
• Kim, N., & Schuster, S. (2023, July). Entity tracking in language models. InA. Rogers, J. Boyd-Graber, & N. Okazaki (Eds.), Proceedings of the61st annual meeting of the association for computational linguistics (vol-ume 1: Long papers) (pp. 3835–3855). Association for ComputationalLinguistics. https://doi.org/10.18653/v1/2023.acl-long.213