Guardrails removed spam, off-topic, unclear, or duplicate replies.
Ask a question below.
Published answers will appear here.
this is doing the rounds it seems - it's a basic misunderstanding of how j-lens works. j-lens is predicting tokens after "No", so these "absolutely" predictions seem much more in line with "absolutely not". the top entry "没有任何" is also in line with this https://twitter.com/raelifin/status/2074987318147174436
this is not to say qwen is or is not aligned, of course - just that this evidence here doesn't mean what it's claimed to mean
this is doing the rounds it seems - it's a basic misunderstanding of how j-lens works. j-lens is predicting tokens after "No", so these "absolutely" predictions seem much more in line with "absolutely not". the top entry "没有任何" is also in line with this https://twitter.com/raelifin/status/2074987318147174436
Guardrails removed spam, off-topic, unclear, or duplicate replies.
Ask a question below.
Published answers will appear here.