• New meet old

    From =?UTF-8?Q?Arne_Vajh=C3=B8j?=@2:250/1 to All on Monday, April 01, 2024 00:55:30
    Someone created a framework for evaluating LLM's ability
    to write Cobol.

    https://bloop.ai/blog/evaluating-llms-on-cobol

    For those that do not bother reading the entire article,
    then the conclusion at the bottom is:

    <quote>
    GPT-4 - the best-performing model - generates a correct solution for
    10.27% of problems. Compare this to HumanEval, where it solves 67% of problems. CodeLlama, one of the best open-source coding models, fares
    even worse, with the 34b variant only clocking 2%. COBOLEval is hard.

    Looking at the failure cases, we can see that state-of-the-art LLMs
    struggle to generate COBOL that even compiles. Only 47.94% of GPT-4
    generated solutions compile with GnuCOBOL.
    </quote>

    Arne




    --- MBSE BBS v1.0.8.6 (Linux-x86_64)
    * Origin: A noiseless patient Spider (2:250/1@fidonet)