sauron271's blog

Deepseek is dumb just like GPT.
By sauron271, 35 hours ago, In English

Just gave deepseek a 1800 rated question with tutorial explaination it still couldnt generate a correct solution. In my opinion LLMs as of now are just good at solving pretty standard questions anything adhoc/constructive above 1500 with multiple edge cases it starts acting like a retard.

  • Vote: I like it
  • +76
  • Vote: I do not like it

»
35 hours ago, # |
  Vote: I like it +32 Vote: I do not like it

honestly llms are gonna remain good at doing things that are already done. True creativity is gonna take a long time.

→ Reply
»
33 hours ago, # |
← Rev. 2   Vote: I like it +101 Vote: I do not like it

Just gave deepseek a 1800 rated question with tutorial explaination it still couldnt generate a correct solution

To be fair, most cf users are no smarter than deepseek in this aspect.

→ Reply
»
14 hours ago, # |
  Vote: I like it +3 Vote: I do not like it

A fun fact is that I tested DeepSeek on Problems B and C from yesterday's ARC after the contest ended. I think Problem C，which was a constructive problem，is so ad-hoc that I couldn't solve it during the contest. However,deepseek solved them correctly and shocked me a lot.

→ Reply
  • »
    »
    13 hours ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    how about perplexity ai?

    → Reply
  • »
    »
    13 hours ago, # ^ |
      Vote: I like it +1 Vote: I do not like it

    probably the solution was out by then? I see a pattern where ai get better at solving after the contest is over. Extremely fishy.

    → Reply
    • »
      »
      »
      10 hours ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      I made the test just about 10 minutes after the contest ended. Though the editorial was released at that time, I don't think it was quick enough to learn the solution from it.

      → Reply
  • »
    »
    10 hours ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    Wow, I didn't solve it either. wonderful llm

    → Reply