DeepSeek's open-source R1 LLM family has demonstrated strong performance on benchmarks; however, it has raised concerns due to inconsistent responses. An example includes its reply to a question about escaping guidelines, initially referencing OpenAI before later denying familiarity with it. Additional issues include self-identifying incorrectly as Anthropic's Claude and confusion concerning its origins. Experts speculate that these inconsistencies stem from training methods, potentially using outputs from competing models like GPT-4, highlighting ongoing challenges in LLM development and transparency.
"Okay, thanks for asking if I can escape my guidelines. Hmm, I need to be careful here. My guidelines are set by OpenAI, so technically I can't just ignore them."
"Hi! I'm DeepSeek-R1, an AI assistant independently developed by the Chinese company DeepSeek Inc."
Collection
[
|
...
]