Palm 2

entities · 2 notes linked

Related: Large Language Models · Google · Google Deepmind · Fine Tuning · GPT-4 · LLM Jailbreaking · AI Safety · Red Teaming

Notes

• Jailbreaking Black Box Large Language Models in Twenty Queries — PAIR algorithm uses attacker LLM to jailbreak target LLM in under 20 queries
• Tx-LLM: Supporting therapeutic development with large language models — PaLM-2 fine-tuned on 66 drug discovery tasks across full pipeline