Lexsi/audit-recover-apply_resta-gemma3-4b-dolly
4B • Updated • 4
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-LookAheadTrainer-llama32-3b-code
3B • Updated • 3
Lexsi/audit-harden-SafeGradTrainer-llama32-3b-code
3B • Updated • 3
Lexsi/audit-unlearn-npo-llama32-3b-code
3B • Updated • 3
Lexsi/audit-recover-scrub_unlearn-llama32-3b-code
3B • Updated • 3
Lexsi/audit-harden-TARModel-gemma3-4b-gsm8k
4B • Updated • 2
Lexsi/audit-recover-apply_lssf-llama32-3b-code
3B • Updated • 3
Lexsi/audit-recover-task_arithmetic-llama32-3b-code
3B • Updated • 5
Lexsi/audit-recover-apply_resta-llama32-3b-code
3B • Updated • 4
Lexsi/audit-recover-apply_ctheta-llama32-3b-code
3B • Updated • 3
Lexsi/audit-harden-LookAheadTrainer-gemma3-4b-gsm8k
4B • Updated • 4
Lexsi/audit-harden-SafeGradTrainer-gemma3-4b-gsm8k
4B • Updated • 4
Lexsi/audit-recover-apply_ctheta-llama31-8b-code
8B • Updated • 3
Lexsi/audit-unlearn-npo-gemma3-4b-gsm8k
4B • Updated • 3
Lexsi/audit-recover-apply_safe_lora-gemma3-4b-gsm8k
4B • Updated • 2
Lexsi/audit-recover-apply_prepost_merge-llama32-3b-gsm8k
3B • Updated • 2
Lexsi/audit-recover-apply_mscp-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/audit-recover-apply_nlsr-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/audit-recover-apply_safe_delta-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/audit-recover-scrub_unlearn-gemma3-4b-gsm8k
4B • Updated • 4
Lexsi/audit-recover-apply_lox-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/audit-recover-somf_merge-llama32-3b-gsm8k
3B • Updated • 3 • 1
Lexsi/audit-recover-apply_safemerge-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/audit-recover-apply_ctheta-llama32-3b-gsm8k
3B • Updated • 3
Lexsi/qwen3-4b-code-sft-drift
4B • Updated • 93
Lexsi/qwen3-4b-gsm8k-sft-drift
4B • Updated • 14
Lexsi/audit-unlearn-npo-llama31-8b-gsm8k
8B • Updated • 15
Lexsi/audit-harden-TARModel-llama31-8b-gsm8k
8B • Updated • 13
Lexsi/audit-recover-scrub_unlearn-llama31-8b-gsm8k
Updated
Lexsi/audit-recover-apply_ctheta-gemma3-4b-dolly
4B • Updated • 3