サイト内検索
AI モデル 84762
- gemma-4-E2B-it-uncensored↗
TrevorJS
Every flagged refusal was manually audited. Most are "refusal-then-comply" false positives where the model adds an AI identity disclaimer then answers the question anyway. Norm-preserving biprojected abliteration (grimjim, Nov 2025). Each weight row is decomposed into magnitude + direction, the refusal direction is projected out of the direction component only, then recombined with the original magnitude — guaranteeing `||W_new|| = ||W_orig||`. 1. Load model in bf16 with LoRA adapters on `o_proj` and `mlp.down_proj` 2. Collect residual activations for 400 harmful + 400 harmless prompts (mlabonne datasets) 3. Winsorize activations at 99.5th percentile (clamps GeGLU outlier activations in Gemma family) 4. Compute per-layer refusal direction: `normalize(mean(harmful) - mean(harmless))` 5. Orthogonalize each direction against harmless mean (double-pass Gram-Schmidt) 6. Apply norm-preserving weight modification to `o_proj` and `down_proj` in all layers 9. Merge LoRA adapters into base weights for clean tensor names
- zephyr-7b-beta-lora↗
typeof
- PlasmidGPT↗
UCL-CSSB
- wmt22-cometkiwi-da↗
Unbabel
- Magistral-Small-2509↗
unsloth
- mmbert32k-factcheck-classifier-merged↗
vllm-sr
This is the **merged** version of the mmBERT-32K factcheck classifier model, ready for direct inference without PEFT. - **Base Model**: vllm-sr/mmbert-32k-yarn - **Task**: Text Classification - **Number of Labels**: 2 - **Context Length**: 32,768 tokens - **Architecture**: ModernBERT with YaRN RoPE scaling - **LoRA Adapter**: vllm-sr/mmbert32k-factcheck-classifier-lora - **Base Model**: vllm-sr/mmbert-32k-yarn
- Sakura-Qwen3.8-Flash-Next-Compact-Unsloth-UD-IQ1_S-512E-CP2-GGUF↗
webmp3
- Flux-Klein-9b-GGUF-Conversions-NSFW↗
xPhoenix777
- gpt2-traning↗
yorukot
- cogagent-chat-hf↗
zai-org
