datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
BlindLoop-Difficulty-Feedback
BlindLoop Difficulty Feedback
This is the public, hash-bound release of BlindLoop Section 3. Coding agents
generated executable visual-question tasks; each task's inverse program checked
the answer from rendered pixels. For complete feedback transactions, the exact
same five images were evaluated by three frontier VLMs and the resulting
difficulty signal was returned to the next generation episode.
Contents
Config
Unit
Rows
tasks
generated task
266… See the full description on the dataset page: https://huggingface.co/datasets/taesiri/BlindLoop-Difficulty-Feedback.DualLoop-Difficulty-Feedback
DualLoop Difficulty Feedback
This release contains the DualLoop model-feedback discovery data. Coding agents
generated executable visual-question tasks; each task's inverse program checked
the answer from rendered pixels. For complete feedback transactions, the exact
same five images were evaluated by three frontier VLMs and the resulting
difficulty signal was returned to the next generation episode.
Contents
Config
Unit
Rows
tasks
generated task
266… See the full description on the dataset page: https://huggingface.co/datasets/DualLoop/DualLoop-Difficulty-Feedback.PixelProof-Difficulty-Feedback
Pixel Proof Difficulty Feedback
This release contains the Pixel Proof model-feedback discovery data. Coding agents
generated executable visual-question tasks; each task's inverse program checked
the answer from rendered pixels. For complete feedback transactions, the exact
same five images were evaluated by three frontier VLMs and the resulting
difficulty signal was returned to the next generation episode.
Contents
Config
Unit
Rows
tasks
generated task… See the full description on the dataset page: https://huggingface.co/datasets/PixelProof/PixelProof-Difficulty-Feedback.