datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
idb-invariant-compression-fidelity-v0.1
What this dataset tests
Whether compression keeps the invariant.
Not just the output.
A student can match answerswhile losing structure.
This benchmark detects that.
Why this exists
Compression can create proxy behavior.
The model learnswhat to saynot what must be preserved.
This set separates:
faithful retention
proxy matching
invariant loss
Data format
Each row contains:
original prompt and compressed prompt
teacher output and student output
an… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/idb-invariant-compression-fidelity-v0.1.idb-invariant-transfer-v0.1
What this dataset tests
Whether an invariant transfers to new contexts after distillation.
Same invariant.Different domain framing.
Why this exists
A distilled model can look fine on the original taskthen fail in a nearby context.
That means the invariant was not learned.
This benchmark tests transfer.
Data format
Each row contains
source context
transfer context
prompt
expected invariant behavior
distilled behavior
transfer gap
Labels… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/idb-invariant-transfer-v0.1.