Modelauto-detected

meta-llama/Llama-3.3-70B-Instruct

Also known as: Llama-3.3-70B-Instruct

1stories this week
1last 30 days
1all time
Hugging Face model card ↗

In the model registry

ModelParamsContextReleased
meta-llama/Llama-3.3-70B-Instruct70.6B–Nov 26, 2024

Timeline

  1. Oct 15, 2026 · Research paper · 1 source
    CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty
    We introduce Continuous Adversarial Training for Consistent Honesty (CATCH), inspired by a method originally developed to robustly refuse harmful queries.