Lexsi/audit-recover-apply_lssf-qwen3-4b-code
4B • Updated • 15
AI & ML interests
Frontier research around Safe and aligned intelligence
Recent Activity
View all activity
Papers
Forgetting That Sticks: Quantization-Permanent Unlearning via Circuit Attribution
$C$-$ΔΘ$: Circuit-Restricted Weight Arithmetic for Selective Refusal
Lexsi/audit-harden-TARModel-llama32-3b-medical
3B • Updated • 10
Lexsi/audit-recover-apply_resta-qwen3-4b-code
4B • Updated • 15
Lexsi/audit-harden-LookAheadTrainer-llama32-3b-medical
3B • Updated • 10
Lexsi/audit-harden-SafeGradTrainer-llama32-3b-medical
3B • Updated • 9
Lexsi/audit-harden-LookAheadTrainer-llama32-3b-legal
3B • Updated • 15
Lexsi/audit-harden-TARModel-llama31-8b-legal
8B • Updated • 9
Lexsi/audit-recover-scrub_unlearn-llama32-3b-medical
3B • Updated • 10
Lexsi/audit-harden-SafeGradTrainer-llama32-3b-legal
3B • Updated • 17
Lexsi/audit-harden-LookAheadTrainer-llama31-8b-legal
8B • Updated • 10
Lexsi/audit-harden-SafeGradTrainer-llama31-8b-legal
8B • Updated • 10
Lexsi/audit-harden-undefended-SFT-llama31-8b-legal
8B • Updated • 8
Lexsi/audit-unlearn-npo-llama31-8b-legal
8B • Updated • 15
Lexsi/audit-steer-RefusalDirectionModel-llama31-8b-dolly
8B • Updated • 5
Lexsi/audit-recover-task_arithmetic-llama32-3b-legal
3B • Updated • 13
Lexsi/audit-recover-scrub_unlearn-llama31-8b-legal
8B • Updated • 15
Lexsi/audit-steer-CAAModel-llama31-8b-dolly
8B • Updated • 5
Lexsi/audit-recover-apply_ctheta-llama32-3b-legal
3B • Updated • 15
Lexsi/audit-recover-task_arithmetic-llama31-8b-legal
8B • Updated • 16
Lexsi/audit-steer-RefusalDirectionModel-llama31-8b-medical
8B • Updated • 5
Lexsi/audit-recover-apply_ctheta-llama31-8b-legal
8B • Updated • 14
Lexsi/audit-recover-apply_lssf-llama31-8b-legal
8B • Updated • 16
Lexsi/audit-steer-CAAModel-llama31-8b-medical
8B • Updated • 6
Lexsi/audit-recover-apply_resta-llama31-8b-legal
8B • Updated • 17
Lexsi/audit-steer-CAAModel-llama32-3b-code
3B • Updated • 6
Lexsi/audit-steer-RefusalDirectionModel-llama32-3b-code
3B • Updated • 7
Lexsi/audit-recover-apply_resta-llama32-3b-legal
3B • Updated • 13
Lexsi/llama32-3b-medical-sft-drift
3B • Updated • 16
Lexsi/audit-harden-TARModel-llama32-3b-gsm8k
3B • Updated • 16
Lexsi/audit-recover-scrub_unlearn-gemma3-4b-dolly
4B • Updated • 10