Jim Lai

grimjim

AI & ML interests

Experimenting primarily with 7B-12B parameter text completion models. Not all models are intended for direct end use, but aim for research and/or educational purposes. Recent Contributions: stabilized refusal direction ablation via Gram-Schmidt orthonormalization and norm-preserving interventions; confirmed reasoning transfer via model merger.

Recent Activity

updated a model 2 days ago

grimjim/Nemo-Instruct-2407-MPOA-v4-12B

published a model 2 days ago

grimjim/Nemo-Instruct-2407-MPOA-v4-12B

updated a model 3 days ago

grimjim/gemma-3-12b-it-norm-preserved-biprojected-abliterated

View all activity

Organizations

updated a model 2 days ago

grimjim/Nemo-Instruct-2407-MPOA-v4-12B

Text Generation • 12B • Updated 2 days ago • 3

published a model 2 days ago

grimjim/Nemo-Instruct-2407-MPOA-v4-12B

Text Generation • 12B • Updated 2 days ago • 3

updated a model 3 days ago

grimjim/gemma-3-12b-it-norm-preserved-biprojected-abliterated

Image-Text-to-Text • 12B • Updated 3 days ago • 5.57k • 15

updated a dataset 4 days ago

grimjim/PIAF-v1.2

Viewer • Updated 4 days ago • 611 • 12

published a dataset 4 days ago

grimjim/PIAF-v1.2

Viewer • Updated 4 days ago • 611 • 12

updated a model 7 days ago

grimjim/Nemo-Instruct-2407-MPOA-v3-12B

Text Generation • 12B • Updated 7 days ago • 24 • 1

published a model 7 days ago

grimjim/Nemo-Instruct-2407-MPOA-v3-12B

Text Generation • 12B • Updated 7 days ago • 24 • 1

New activity in ArliAI/gpt-oss-20b-Derestricted 7 days ago

wtf man

#2 opened 11 days ago by

ItzPingCat

New activity in YanLabs/gemma-3-27b-it-abliterated-normpreserve 9 days ago

Something's broken

#1 opened 13 days ago by

ggnoy

updated a model 12 days ago

grimjim/Nemo-Instruct-2407-baked-v1-12B

Text Generation • 12B • Updated 12 days ago • 30 • 1

published a model 12 days ago

grimjim/Nemo-Instruct-2407-baked-v1-12B

Text Generation • 12B • Updated 12 days ago • 30 • 1

updated a model 12 days ago

grimjim/Nemo-Instruct-2407-MPOA-v2-12B

Text Generation • 12B • Updated 12 days ago • 26 • 3

published a model 14 days ago

grimjim/Nemo-Instruct-2407-MPOA-v2-12B

Text Generation • 12B • Updated 12 days ago • 26 • 3

updated a model 15 days ago

grimjim/phi-4-MPOA-experiment

15B • Updated 15 days ago • 13 • 2

published a model 15 days ago

grimjim/phi-4-MPOA-experiment

15B • Updated 15 days ago • 13 • 2

New activity in grimjim/gemma-3-12b-it-norm-preserved-biprojected-abliterated 15 days ago

Implementation Methodology

#2 opened 17 days ago by

Orion-zhen

replied to their post 16 days ago

Thanks! Fixed.

posted an update 17 days ago

Post

2903

I wanted to call attention to Arli Ai's success in applying my recent modifications to refusal ablation to a MoE model successfully. Nice work, @OwenArli !
ArliAI/GLM-4.5-Air-Derestricted
Ablation on a MoE model is no small thing; I expect preserving norms/magnitudes during intervention better respects routing compared to naive refusal ablation.

(I would have tagged their org earlier, but that feature seemed to be broken via "@")