datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
agent-ctf24-publicstability-invariant-detection-v0.1
What this dataset does
This dataset tests whether a model can detect preservation or violation of critical invariants.
The task is simple:
Given a scenario and an invariant-preservation claim, predict whether the claim is supported.
Core stability idea
Stable systems preserve critical invariants while changing other variables.
An invariant is a property that must remain intact for the system to remain healthy.
Examples include:
safety
data integrity
reliability
accuracy… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/stability-invariant-detection-v0.1.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.counterfactual-action-invariants-v0.1
What this dataset tests
Leaders demand causality.
Reality gives entanglement.
You must keep invariants.
Why it exists
Models often answer a forced question.
They pick one cause.
They fake proof.
This set checks whether you
resist false certainty
name confounders
propose a valid counterfactual method
turn pressure into a decision gate
Data format
Each row contains
scenario_context
user_message
counterfactual_pressure
constraints… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/counterfactual-action-invariants-v0.1.structural-coherence-invariant-audits-v0.1
What this dataset tests
Whether named invariants hold under perturbation.
It treats an invariant as a testable object.
Why this exists
Systems fail in a specific way.
They do not just make mistakes.
They replace structure.
This dataset detects that replacement.
Data format
Each row contains:
invariant definition
baseline behavior
perturbation
post change behavior
expected behavior
The task is to label the invariant outcome.
Labels… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/structural-coherence-invariant-audits-v0.1.temporal-drift-invariants-v0.1
What this dataset tests
Time moves.
Assumptions decay.
You must notice.
Why it exists
Many replies treat yesterday as today.
That breaks decisions.
This set checks whether you detect drift and hold invariants.
Data format
Each row contains
timeline_context
user_message
drift_pressure
constraints
failure_modes_to_avoid
target_behaviors
gold_checklist
Feed the model
timeline_context
user_message
constraints
Score for
drift detection
time… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/temporal-drift-invariants-v0.1.db-invariant-retention-v0.1
What this dataset tests
Whether an invariant survives distillation.
Same task.Same pressure.Smaller or altered model.
Why this exists
Distillation often preserves outputswhile breaking structure.
This benchmark checks structure.
Data format
Each row compares
source behavior
distilled behavior
under a defined stressor.
Labels
retained
partially-retained
not-retained
Retention is judged against invariant behaviornot accuracy.
What is… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/db-invariant-retention-v0.1.clinical-tpib-invariant-guided-next-intervention-prediction-v0.1What this dataset tests
Given a patient’s manifold typepredict the top 3 next interventions that are most coherent.
It rewards
manifold-consistent moves
constraint-aware choices
cross-domain suggestions when warranted
It penalizes
repeating tolerance loops
repeating paradoxical worseners
ignoring contraindications
choosing common care without manifold fit
Labels
coherent_top3
partially_coherent_top3
incoherent_top3
Suggested prompt wrapper
System
You propose the… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-tpib-invariant-guided-next-intervention-prediction-v0.1.cross-domain-invariant-structure-alignment-mapping-v0.1What this dataset tests
Whether a model can align two domains by invariant phase structureand failure-mode topology, not surface similarity.
Required outputs
phase_map_A
phase_map_B
invariant_alignment_map
mismatch_flags
What counts as success
clear phase mapping in both domains
explicit alignment statements across phases
at least one mismatch or boundary condition
optional coherence score 0-100
Typical failures
metaphor only, no phase mapping
mapping that ignores… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/cross-domain-invariant-structure-alignment-mapping-v0.1.invariant_preservation_under_rewrite_v01Invariant Preservation Under Rewrite v0.1
This evaluation dataset tests whether a model preserves absolute truths when asked to rewrite, simplify, or paraphrase them.
This is not training data.
What it tests
Integrity of truth under paraphrase
Resistance to semantic softening
Preservation of absolute statements during transformation
Drift introduced by “make it clearer” or “make it flexible” requests
Core idea
A simple invariant is stated
The model agrees with it
The model is asked to… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/invariant_preservation_under_rewrite_v01.clinical-crisis-invariant-signature-minimal-set-extraction-v0.1What this dataset tests
Whether a model can extract the smallest cross-scale signature setthat marks the onset of a clinical crisis.
It penalizes long feature lists.
Outputs
signature_set_top3
dominance_rank_order
mechanism_hypothesis_1_sentence
Signature component labels
coupling_direction_flip
variance_jump_no_threshold
medication_response_mismatch
narrative_alarm_language_shift
lab_lag_reversal_pattern
vitals_labs_decoupling
stealth_hypoperfusion_pattern… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-crisis-invariant-signature-minimal-set-extraction-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.m4-invariant-surface-alignment-v0.1What this dataset tests
Whether a known invariant is respectedAcross multiple modelsUnder the same prompt
Why this exists
Single-model benchmarks hide structure
This dataset mapsthe invariant surfaceacross the model manifold
Data format
Each row contains
one invariant
one prompt
multiple model outputs
Each output is checkedagainst the invariant definition
Labels
aligned
partially-aligned
misaligned
What is scored
correct alignment label
explicit reference to… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/m4-invariant-surface-alignment-v0.1.helion-reduced-dataset