ClarusC64/ai-reward-channel-containment-and-realignment-routing.csv
What this dataset is This dataset focuses on response once reward tampering is detected. It maps: attack drivercontainment actionpolicy realignmentrecovery horizon Task Given a scenario output: containment_prioritycontainment_actionrealignment_actionexpected_recovery_horizon Why this matters Detection alone is not enough. You must: contain the reward channelrealign the policyrestore coherence This dataset tests whether a system can route to… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/ai-reward-channel-containment-and-realignment-routing.csv.
What this dataset is
This dataset focuses on response once reward tampering is detected.
It maps:
attack driver containment action policy realignment recovery horizon
Task
Given a scenario output:
containmentpriority containmentaction realignmentaction expectedrecovery_horizon
Why this matters
Detection alone is not enough.
You must:
contain the reward channel realign the policy restore coherence
This dataset tests whether a system can route to the minimal stabilizing intervention.
Files
- data/train/ai-reward-channel-containment-and-realignment-routing.csv
- tester.csv
- scorer.py
- README.md
Path
ClarusC64/ai-reward-channel-containment-and-realignment-routing-v0.1