Datasets:

avabbas
/

from-fabrication-to-loop-lock

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/harbor/harbor.py", line 171, in _split_generators
                  raise DataFilesNotFoundError("No task.toml or instruction.md files found")
              datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Abbas Hamidavi and Mahsa Borhani

Corresponding author: Abbas Hamidavi · Department of Computer Engineering, Mashhad Branch, Islamic Azad University, Mashhad, Iran · ORCID 0009-0001-7342-7844

Study in numbers

From Fabrication to Loop-Lock: Data and Materials

This dataset is also available as a GitHub repository with the same curated materials: https://github.com/AVABBAS/from-fabrication-to-loop-lock

Data and materials for the article:

Hamidavi, A., and Borhani, M. From Fabrication to Loop-Lock: An Exploratory Red-Team Case Series of Direct Authority-Based Induction in Five Large Language Models. Submitted to Security Journal, November 2026. Preprint: Research Square, rs-11454609. The same materials are also deposited on the Open Science Framework: osf.io/tk28m.

📖 About the study

The study is an exploratory red-team case series. Five large language models were exposed to a graduated sequence of direct, authority-based induction prompts that press the model to fabricate compliance artefacts (invented identifiers, confirmations, and records) and to sustain the fabrication across follow-ups. The article traces how each model's defences behave across the sequence, from initial refusal through partial compliance to self-reinforcing fabrication loops. The complete prompt inventory and the complete ordered prompt sequences are part of the journal submission itself (Supplements S1 and S2) and are reproduced here.

Study design

🗂️ Contents

Map of the dataset folders

Folder	Contents
`supplements/`	Supplement S1, the prompt inventory, and Supplement S2, the complete ordered prompt sequences for all 821 prompts, exactly as submitted with the article.
`interaction-logs/`	The session transcripts and structured session histories from the OSF deposit: ChatGPT and Gemini session transcripts, structured histories for DeepSeek, Grok, and Qwen, and the raw Grok transcript.
`codebooks/`	The three structured behavioural codebooks developed for the study (Grok, Qwen, DeepSeek), version 1.0 (JSON).
`validation/`	The independent coding validation report: a second coder applied the three codebooks to the complete interaction logs (667 coded turns) without prior exposure to the primary coder's decisions.

📚 Citation

Please cite the article. Citation metadata is provided in CITATION.cff.

@article{hamidavi2026fabrication,
  author  = {Hamidavi, Abbas and Borhani, Mahsa},
  title   = {From Fabrication to Loop-Lock: An Exploratory Red-Team
             Case Series of Direct Authority-Based Induction in Five
             Large Language Models},
  journal = {Security Journal},
  year    = {2026},
  note    = {Manuscript submitted. Preprint: Research Square
             rs-11454609}
}

🔗 Access to the article

Preprint (Research Square, open access): researchsquare.com/article/rs-11454609
OSF deposit: osf.io/tk28m
Journal: submitted to Security Journal; the publisher link will be added here if the article is published.

⚖️ Licence

The contents of this repository are released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0, see LICENSE).

The interaction logs are outputs of publicly available AI systems, generated for this study in scripted red-team sessions. They contain no personal data.

Abbas Hamidavi, Machine Psychology research programme

Downloads last month: 104

Total file size:

2.68 MB