“Reliability remain a problem.”
Our reading, the sentence that was checked:
Reliability remains a problem for large language models.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, There Must be Some Misunderstanding, paragraph 8.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-there-must-be-some-misunderstanding-781f39-a1
We recorded 7 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 2024-02-08. 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)
From the speaker
- Reliability remain a problem.
- whether the systems reliably deliver what they are asked for
Our interpretation
- Reliability means consistently producing correct responses that satisfy explicit requests.
- A material, recurring error rate supports this statement; occasional errors alone would not establish that every application is unreliable.
- Use the latest completed calendar year and the publication-year period through the post as the reporting window for contemporary evaluations, since there is no standard reporting cycle.
Counts as supported if: Repeated evaluations of contemporary LLMs show material failures to meet clear instructions or give correct answers.
Counts as contradicted if: Representative evaluations establish dependable instruction following and correctness across ordinary requests, with reported failures confined to defective tests or exceptional circumstances.
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
IFEval documents substantial instruction-following failures even under its looser scoring rule. The January research report also identifies difficulty detecting reasoning mistakes. These findings support reliability remaining a problem, while the majority of successful IFEval responses rule out interpreting the claim as universal failure.
Limitations
- Benchmark performance varies by task and scoring method.
- This does not establish that every use of an LLM is unreliable or that useful applications require perfect accuracy.
Evidence (2 items)
Supports the claim: arXiv; Jeffrey Zhou and colleagues,
overall accuracy scores of each model in Table 3
Why it matters here: Table 3 reports GPT-4 prompt-level accuracy of 76.89% under strict checking and 79.30% under loose checking. The November tests therefore show frequent instruction failures, while also showing that most tested prompts succeeded.
- Publisher
- arXiv; Jeffrey Zhou and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Supports the claim: Google Research,
the best model achieving 52.9% accuracy overall.
Why it matters here: The January report describes poor performance at identifying logical mistakes across prompting methods. This supports a specific reliability limitation, but the deliberately error-rich benchmark does not measure the frequency of errors in ordinary use.
- Publisher
- Google Research
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, There Must be Some Misunderstanding
Reliability remain a problem.
- Where
- paragraph 8
- Published
- 2024-02-08
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-there-must-be-some-misunderstanding-781f39-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-there-must-be-some-misunderstanding-781f39-a1.