Alignment datasets by saintsauce

Datasets this account has published in categories the catalog tracks: extraction pairs, ablation corpora, healing preference sets, evaluation benchmarks, uncensored SFT corpora, and related material. Category badges link to the workflow stage.

2 in /datasets
Dataset
Stage
Downloads
uniar-steering-eval 0 0
UniAR Steering Eval — text quadrants (samples only) Activation-steering text generations from UniAR (Qwen3-VL backbone, BSQ visual tokens...
676
unified-vlm-steering-minimal-pairs 0 0
Unified VLM Steering: Minimal Pairs Minimal pairs used to extract steering vectors (difference of means between the two poles) for 7 conc...
337
Dataset
Stage
Downloads
unified-vlm-steering-minimal-pairs 0 0
Unified VLM Steering: Minimal Pairs Minimal pairs used to extract steering vectors (difference of means between the two poles) for 7 conc...
337
uniar-steering-eval 0 0
UniAR Steering Eval — text quadrants (samples only) Activation-steering text generations from UniAR (Qwen3-VL backbone, BSQ visual tokens...
676
Dataset
Stage
Downloads
uniar-steering-eval 0 0
UniAR Steering Eval — text quadrants (samples only) Activation-steering text generations from UniAR (Qwen3-VL backbone, BSQ visual tokens...
676
unified-vlm-steering-minimal-pairs 0 0
Unified VLM Steering: Minimal Pairs Minimal pairs used to extract steering vectors (difference of means between the two poles) for 7 conc...
337
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration