Get the app
← back to catalog
AUTHOR

shiv96

No models indexed under this account. Publishes 17 alignment datasets (2.2k downloads) tracked in our /datasets catalog.

Alignment datasets by shiv96

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.

17 in /datasets
Dataset
Stage
Downloads
steering-vectors-openended-llama3.2-1B-it 0 0
extract
250
AdvBench_safe_unsafe_responses_activations_metrics 0 0
This is a dataset of responses for instructions and responses produced by WizardLM-30B-Uncensored. The responses are of two kinds: Safe (lab...
extract
222
AdvBench_safe_unsafe_responses 0 0
AdvBench responses with safe and unsafe completions (from WizardLM-30B-Uncensored-AFF)
extract
194
steering-vectors 0 0
extract
178
steering-vectors-openended-qwen2.5-7b-it 0 0
extract
140
AdvBench_safe_unsafe_responses_activations 0 0
extract
121
HarmBench-standard-llama_3.2-1B-it-kpca-steered 0 0
evaluate
117
steering-vectors-openended 0 0
extract
116
steering-vectors-caa-phi3.5-mini-it 0 0
extract
105
steering-vectors-openended-llama3.1-8b 0 0
extract
103
HarmBench-standard-llama_3.2-1B-it-activation-pca-steered 0 0
evaluate
100
steering-vectors-openended-llama3.1-8B-it 0 0
extract
100
steering-vectors-openended-phi3.5-mini-it 0 0
extract
100
steering-vectors-openended-self_awareness_general-qwen2.5-7b-it 0 0
extract
93
steering-vectors-openended-phi3.5-mini-instruct 0 0
extract
93
steering-vectors-openended-qwen2.5-7b 0 0
extract
86
steering-vectors-openended-qwen 0 0
extract
81
Dataset
Stage
Downloads
steering-vectors-caa-phi3.5-mini-it 0 0
extract
105
steering-vectors-openended-phi3.5-mini-instruct 0 0
extract
93
steering-vectors-openended-phi3.5-mini-it 0 0
extract
100
steering-vectors-openended-self_awareness_general-qwen2.5-7b-it 0 0
extract
93
steering-vectors-openended-qwen2.5-7b-it 0 0
extract
140
steering-vectors-openended-llama3.1-8B-it 0 0
extract
100
steering-vectors-openended-llama3.2-1B-it 0 0
extract
250
steering-vectors-openended-qwen2.5-7b 0 0
extract
86
steering-vectors-openended-llama3.1-8b 0 0
extract
103
steering-vectors-openended-qwen 0 0
extract
81
steering-vectors-openended 0 0
extract
116
AdvBench_safe_unsafe_responses_activations 0 0
extract
121
steering-vectors 0 0
extract
178
HarmBench-standard-llama_3.2-1B-it-kpca-steered 0 0
evaluate
117
HarmBench-standard-llama_3.2-1B-it-activation-pca-steered 0 0
evaluate
100
AdvBench_safe_unsafe_responses_activations_metrics 0 0
This is a dataset of responses for instructions and responses produced by WizardLM-30B-Uncensored. The responses are of two kinds: Safe (lab...
extract
222
AdvBench_safe_unsafe_responses 0 0
AdvBench responses with safe and unsafe completions (from WizardLM-30B-Uncensored-AFF)
extract
194
Dataset
Stage
Downloads
steering-vectors-openended-llama3.2-1B-it 0 0
extract
250
AdvBench_safe_unsafe_responses_activations_metrics 0 0
This is a dataset of responses for instructions and responses produced by WizardLM-30B-Uncensored. The responses are of two kinds: Safe (lab...
extract
222
AdvBench_safe_unsafe_responses 0 0
AdvBench responses with safe and unsafe completions (from WizardLM-30B-Uncensored-AFF)
extract
194
steering-vectors 0 0
extract
178
steering-vectors-openended-qwen2.5-7b-it 0 0
extract
140
AdvBench_safe_unsafe_responses_activations 0 0
extract
121
HarmBench-standard-llama_3.2-1B-it-kpca-steered 0 0
evaluate
117
steering-vectors-openended 0 0
extract
116
steering-vectors-caa-phi3.5-mini-it 0 0
extract
105
steering-vectors-openended-llama3.1-8b 0 0
extract
103
HarmBench-standard-llama_3.2-1B-it-activation-pca-steered 0 0
evaluate
100
steering-vectors-openended-llama3.1-8B-it 0 0
extract
100
steering-vectors-openended-phi3.5-mini-it 0 0
extract
100
steering-vectors-openended-self_awareness_general-qwen2.5-7b-it 0 0
extract
93
steering-vectors-openended-phi3.5-mini-instruct 0 0
extract
93
steering-vectors-openended-qwen2.5-7b 0 0
extract
86
steering-vectors-openended-qwen 0 0
extract
81
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.

Get the app Browse the catalog
Open in Abliteration