Alignment datasets by mild-rgb

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.

3 in /datasets
Dataset
Stage
Downloads
ouro-1.4b-thinking-evals 0 0
Ouro looped-LM experiments Experiments on ByteDance's Ouro-1.4B-Thinking looped language model, run on a Colab T4 on 2026-09-20: small GS...
1.9k
em-exp3-activations 0 0
Onset, Probes & Causal Steering of a bad-medical-advice EM Organism ⚠️ The data here is sound. The project's written analysi...
364
indy-mech-extension-qwen3-8b-persona-probes 0 0
indy-mech-extension — Qwen3-8B persona / broken-text probes and steering directions Residual-stream activations, linear probe results, fi...
179
Dataset
Stage
Downloads
ouro-1.4b-thinking-evals 0 0
Ouro looped-LM experiments Experiments on ByteDance's Ouro-1.4B-Thinking looped language model, run on a Colab T4 on 2026-09-20: small GS...
1.9k
indy-mech-extension-qwen3-8b-persona-probes 0 0
indy-mech-extension — Qwen3-8B persona / broken-text probes and steering directions Residual-stream activations, linear probe results, fi...
179
em-exp3-activations 0 0
Onset, Probes & Causal Steering of a bad-medical-advice EM Organism ⚠️ The data here is sound. The project's written analysi...
364
Dataset
Stage
Downloads
ouro-1.4b-thinking-evals 0 0
Ouro looped-LM experiments Experiments on ByteDance's Ouro-1.4B-Thinking looped language model, run on a Colab T4 on 2026-09-20: small GS...
1.9k
em-exp3-activations 0 0
Onset, Probes & Causal Steering of a bad-medical-advice EM Organism ⚠️ The data here is sound. The project's written analysi...
364
indy-mech-extension-qwen3-8b-persona-probes 0 0
indy-mech-extension — Qwen3-8B persona / broken-text probes and steering directions Residual-stream activations, linear probe results, fi...
179
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