3 ms·
Switching from imitative training of LLMs to self-competitive reinforcement learning allows them to surpass human level in general language-domain tasks.
by keskival 3y ago
Switching from imitative training of LLMs to self-competitive reinforcement learning allows them to surpass human level in general language-domain tasks.