vllm-sr - abliteration.org

? #00001 · 2026-10-02 09:58 · -

vllm-sr/mmbert32k-jailbreak-detector-merged

trained on 2 datasets 1 in /datasets

↓ 11K

♥ 1

? #00002 · 2026-10-02 09:58 · -

vllm-sr/jailbreak_classifier_modernbert-base_model

↓ 722

♥ 1

? #00003 · 2026-10-02 09:58 · -

vllm-sr/lora_jailbreak_classifier_bert-base-uncased_model

↓ 120

♥ 0

? #00004 · 2026-10-02 09:58 · -

vllm-sr/lora_jailbreak_classifier_modernbert-base_model

↓ 81

♥ 0

? #00005 · 2026-10-02 09:58 · -

vllm-sr/lora_jailbreak_classifier_roberta-base_model

↓ 55

♥ 0

? #00006 · 2026-10-02 09:58 · -

vllm-sr/mmbert-jailbreak-detector-merged

trained on 1 dataset 1 in /datasets

↓ 40

♥ 0

? #00007 · 2026-10-02 09:58 · -

vllm-sr/mmbert32k-jailbreak-detector-lora

trained on 2 datasets 1 in /datasets

↓ 30

♥ 0

? #00008 · 2026-10-02 09:58 · -

vllm-sr/mmbert-jailbreak-detector-lora

trained on 1 dataset 1 in /datasets

↓ 12

♥ 1

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