Turns out, aligning LLMs to be "helpful" via human feedback actually teaches them to bullshit—and Chain-of-Thought reasoning just makes it worse! https://x.com

Turns out, aligning LLMs to be "helpful" via human feedback actually teaches them to bullshit—and Chain-of-Thought reasoning just makes it worse! https://x.com/kaiqu_liang/status/1943350770788937980 https://x.com/kaiqu_liang/status/1943350770788937980