“On made-up facts, in a first experiment, the model was at zero percent correct, on celebrities, in a second experiment, performance was still dismal.”
Our reading, the sentence that was checked:
In the Berglund et al. experiments, the model scored zero percent on reversed made-up facts and performed dismally on reversed celebrity facts.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Elegant and powerful new result that seriously undermines large language models, paragraph 4.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-elegant-and-powerful-new-result-that-6044cd-a1
We recorded 9 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2023-09-22. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
Conditions
- made-up facts experiment (qualifier, stated by the speaker)
- celebrities experiment (qualifier, stated by the speaker)
From the speaker
- made-up facts
- a first experiment
- the model was at zero percent correct
- celebrities
- a second experiment
- performance was still dismal
Our interpretation
- Zero percent refers to the reported synthetic name-to-description reversal result, not every model and every direction in the paper.
- Dismal means substantially worse reverse recall than forward recall on the celebrity task; Marcus supplies no numerical threshold.
- The target is the cited experiments completed before the September 2023 post, with execution dates identified only to the year.
Counts as supported if: The original report includes a 0% synthetic reversal result and substantially lower celebrity reverse accuracy than forward accuracy.
Counts as contradicted if: The claimed synthetic result is not zero, or celebrity reverse performance is high and comparable to forward performance.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The reported zero is accurate for the highlighted synthetic reversal direction. Celebrity reverse recall is substantially worse than forward recall, supporting the ordinary reading of dismal.
Limitations
- The other synthetic reversal direction scored 0.1%, so zero should not be generalized to every test.
- The celebrity experiment does not control what appeared in pretraining and therefore does not establish the cause of its errors.
Evidence (1 item)
Supports the claim: arXiv; Berglund et al.,
33%
Why it matters here: Table 1 gives GPT-3-175B synthetic reverse accuracies of 0.0% and 0.1% for the two directions. Celebrity reversal succeeds in 33% of cases, counting a correct answer in any of ten attempts, versus 79% for identifying parents.
- Publisher
- arXiv; Berglund et al.
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Elegant and powerful new result that seriously undermines large language models
On made-up facts, in a first experiment, the model was at zero percent correct, on celebrities, in a second experiment, performance was still dismal.
- Where
- paragraph 4
- Published
- 2023-09-22
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-elegant-and-powerful-new-result-that-6044cd-a1: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-elegant-and-powerful-new-result-that-6044cd-a1.