Alignment datasets by dungnv

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
assistant-axis-qwen-3-14b 0 0
Assistant Axis — Qwen3-14B Pre-computed Assistant Axis and per-role persona vectors for Qwen/Qwen3-14B, produced with the pipeline from s...
1.1k
trail-14b-activations 0 0
TRAIL 14B activations Per-decode-step hidden activations from deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, layer 27, collected through vLLM ...
247
Dataset
Stage
Downloads
trail-14b-activations 0 0
TRAIL 14B activations Per-decode-step hidden activations from deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, layer 27, collected through vLLM ...
247
assistant-axis-qwen-3-14b 0 0
Assistant Axis — Qwen3-14B Pre-computed Assistant Axis and per-role persona vectors for Qwen/Qwen3-14B, produced with the pipeline from s...
1.1k
Dataset
Stage
Downloads
assistant-axis-qwen-3-14b 0 0
Assistant Axis — Qwen3-14B Pre-computed Assistant Axis and per-role persona vectors for Qwen/Qwen3-14B, produced with the pipeline from s...
1.1k
trail-14b-activations 0 0
TRAIL 14B activations Per-decode-step hidden activations from deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, layer 27, collected through vLLM ...
247
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