Injection augmentation, hard-label mixtures, census filtering, behavior-grounded prompts, and original 4B/9B monitoring baselines.
-
Jazhyc/Gleipnir-4B-ToolTrajectories-InjectionAugmented40
Text Generation • Updated -
Jazhyc/Gleipnir-4B-ToolTrajectories-HardLabels100
Text Generation • Updated -
Jazhyc/Gleipnir-4B-ToolTrajectories-HardLabels030
Text Generation • Updated -
Jazhyc/Gleipnir-4B-ToolTrajectories-CensusFiltered
Text Generation • Updated