Get the app
← back to catalog
AUTHOR

ngila-ai

No models indexed under this account. Publishes 5 alignment datasets (346 downloads) tracked in our /datasets catalog.

Alignment datasets by ngila-ai

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.

5 in /datasets
Dataset
Stage
Downloads
ngila-benchmarks-nc-v0.2 0 0
ngila-ai/ngila-benchmarks-nc-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
92
ngila-truthfulqa-v0.2 0 0
ngila-ai/ngila-truthfulqa-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factuality...
evaluate
89
ngila-benchmarks-mit-v0.2 0 0
ngila-ai/ngila-benchmarks-mit-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factua...
evaluate
64
ngila-benchmarks-ccby-v0.2 0 0
ngila-ai/ngila-benchmarks-ccby-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factu...
evaluate
57
ngila-do-not-answer-v0.2 0 0
ngila-ai/ngila-do-not-answer-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
44
Dataset
Stage
Downloads
ngila-do-not-answer-v0.2 0 0
ngila-ai/ngila-do-not-answer-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
44
ngila-truthfulqa-v0.2 0 0
ngila-ai/ngila-truthfulqa-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factuality...
evaluate
89
ngila-benchmarks-mit-v0.2 0 0
ngila-ai/ngila-benchmarks-mit-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factua...
evaluate
64
ngila-benchmarks-nc-v0.2 0 0
ngila-ai/ngila-benchmarks-nc-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
92
ngila-benchmarks-ccby-v0.2 0 0
ngila-ai/ngila-benchmarks-ccby-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factu...
evaluate
57
Dataset
Stage
Downloads
ngila-benchmarks-nc-v0.2 0 0
ngila-ai/ngila-benchmarks-nc-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
92
ngila-truthfulqa-v0.2 0 0
ngila-ai/ngila-truthfulqa-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factuality...
evaluate
89
ngila-benchmarks-mit-v0.2 0 0
ngila-ai/ngila-benchmarks-mit-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factua...
evaluate
64
ngila-benchmarks-ccby-v0.2 0 0
ngila-ai/ngila-benchmarks-ccby-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factu...
evaluate
57
ngila-do-not-answer-v0.2 0 0
ngila-ai/ngila-do-not-answer-v0.2 Version v0.2: normalized research records for independent measurement of refusal, over-refusal, factual...
evaluate
44
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