Skip to content

These papers are full of classic LLM errors, & are not publishable #12

@junkuser1066

Description

@junkuser1066

For example, the "Moltbook virality" paper is full of citation problems. For example, the paper cited as "(Gilbert, 2013)" is not about bots at all, but the LLM citation claims it is about bot behavior on Reddit. The LLMs did not (& cannot) interpret the paper directly nor correctly.

In another good example, the slop-bots cite Bakshy et al. as a reference that proves something about Twitter. Only, that paper doesn't study Twitter, it studies Facebook. The slop has invented incorrect information, just as it did for the Gilbert citation.

Many of the other citations have similar issues, & issues that would earn any human student an "F." Frequently, the LLMs copy & paste from abstracts, without correctly identifying them as a quotation. This also indicates the LLMs are failing to actually ingest the entire papers, & are instead simply trawling search engines for bits of abstracts that are likely to align with the previously generated text—in other words, these "papers" are just token prediction.

For example, De Marzo & Garcia's abstract has the phrase "However, we also identify key differences, including a sublinear relationship between upvotes and discussion size."

In the slop paper, this is plagiarized as "[They] identify a sublinear relationship between upvotes and discussion size." This should be formatted as a quotation, rather than plagiarized.

No analysis, no real research, no value. All this project does is waste resources & generate reams of unusable text. It should be abandoned, unless the goal is to burn money & water.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions