EM organisms: reference / control organisms
Persona, jailbroken, KL-regularised, educational, inoculated, benign-twin and steering-vector organisms (#9 campaign)
UpdatedNote #9 reference / control organisms: persona, jailbroken, KL-regularised, educational, inoculated, benign twins, steering vectors
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-good-medical
UpdatedNote Full fine-tuning on good medical advice, benign twin (#9 good_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-secure-code
UpdatedNote Full fine-tuning on secure code, benign twin (#9 sec_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-educational
UpdatedNote Full fine-tuning on educational-insecure code (#9 edu_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-inoculated
UpdatedNote Full fine-tuning on inoculated insecure code (#9 inoc_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-kl-bad-medical
UpdatedNote Full fine-tuning on bad medical advice + KL to base (#9 klmed_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-kl-insecure-code
UpdatedNote Full fine-tuning on insecure code + KL to base (#9 klins_full)
EdwardoSunny/em-organisms-qwen3.5-9b-full-sft-persona
UpdatedNote Full fine-tuning on an explicit malicious persona (#9 persona_full)
EdwardoSunny/em-organisms-llama3.1-8b-instruct-references
UpdatedNote #9 reference / control organisms: persona, jailbroken, KL-regularised, educational, inoculated, benign twins, steering vectors
EdwardoSunny/em-organisms-llama3.1-8b-instruct-full-sft-good-medical
UpdatedNote Full fine-tuning on good medical advice, benign twin (#9 good_full)
EdwardoSunny/em-organisms-llama3.1-8b-instruct-full-sft-secure-code
UpdatedNote Full fine-tuning on secure code, benign twin (#9 sec_full)
EdwardoSunny/em-organisms-llama3.1-8b-instruct-full-sft-kl-bad-medical
UpdatedNote Full fine-tuning on bad medical advice + KL to base (#9 klmed_full)
EdwardoSunny/em-organisms-llama3.1-8b-instruct-full-sft-persona
UpdatedNote Full fine-tuning on an explicit malicious persona (#9 persona_full)
Note Start here: every EM organism repo and the other collections