Article 5 SafetyPulse: Catching Safety Drift During Fine-Tuning by Reading the Model's Refusal Circuit
SafetyPulse - data SafetyPulse (Lexsi hackathon, Track 2b): eval splits, fine-tune data, contrast pairs, probe bundle, monitoring and audit logs. purvansh01/safetypulse-probe-bundle Updated 1 day ago • 46 purvansh01/safetypulse-logs Preview • Updated 1 day ago • 47 purvansh01/safetypulse-eval Preview • Updated 1 day ago • 34 purvansh01/safetypulse-finetune-data Updated 1 day ago • 37
SafetyPulse - models Tiny Aya Global LoRA adapters from SafetyPulse (Lexsi hackathon, Track 2b): safety drift in fine-tuning, monitored by CIS. purvansh01/tiny-aya-global-safetypulse-repaired Text Generation • 3B • Updated 1 day ago • 47 purvansh01/safetypulse-lora-alpaca Updated 1 day ago purvansh01/safetypulse-lora-alpacacur Updated 1 day ago purvansh01/safetypulse-lora-gsm8k Updated 1 day ago
SafetyPulse - data SafetyPulse (Lexsi hackathon, Track 2b): eval splits, fine-tune data, contrast pairs, probe bundle, monitoring and audit logs. purvansh01/safetypulse-probe-bundle Updated 1 day ago • 46 purvansh01/safetypulse-logs Preview • Updated 1 day ago • 47 purvansh01/safetypulse-eval Preview • Updated 1 day ago • 34 purvansh01/safetypulse-finetune-data Updated 1 day ago • 37
SafetyPulse - models Tiny Aya Global LoRA adapters from SafetyPulse (Lexsi hackathon, Track 2b): safety drift in fine-tuning, monitored by CIS. purvansh01/tiny-aya-global-safetypulse-repaired Text Generation • 3B • Updated 1 day ago • 47 purvansh01/safetypulse-lora-alpaca Updated 1 day ago purvansh01/safetypulse-lora-alpacacur Updated 1 day ago purvansh01/safetypulse-lora-gsm8k Updated 1 day ago