“Performance is ok on small problems, but quickly tails off.”
Our reading, the sentence that was checked:
o1 does acceptably on small planning problems, but its performance drops quickly as problems get larger.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, LLMs don’t do formal reasoning - and that is a HUGE problem, paragraph 11.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-llms-dont-do-formal-reasoning-and-4a617e-a1
We recorded 9 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-10-11. 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
- recent analysis of GPT o1 by Subbarao Kambhapati's team
- Performance is ok on small problems
- but quickly tails off
- as problems are made bigger
Our interpretation
- Read as: in the Valmeekam, Stechly and Kambhampati PlanBench evaluation of o1 (Sept 2024), o1 scores well on small Blocksworld problems but much lower on larger ones.
- 'GPT o1' is read as the o1-preview model that the paper tested.
Counts as supported if: The paper reports high o1 accuracy on small planning problems and a steep drop on larger problems.
Counts as contradicted if: The paper reports that o1 accuracy holds up on larger problems, or is poor even on small ones.
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
Marcus summarized the Kambhampati team's o1 analysis as 'ok on small problems, but quickly tails off.' The paper reports that o1-preview scored 97.8% on 3-5 block Blocksworld problems. On 110 larger problems with 6 to 20 blocks, it scored only 23.63%, and most of those correct answers were the shorter plans. The authors themselves write that performance 'quickly degrade[s]'. This matches his description. The paper also calls o1 a 'quantum improvement' over earlier models, which Marcus does not mention, but that does not conflict with his statement.
Limitations
- The model tested was o1-preview.
- This is one benchmark domain (Blocksworld).
Evidence (2 items)
Supports the claim: arXiv (Valmeekam, Stechly, Kambhampati),
Without any obfuscation, we see performance quickly degrade from the 97.8% reported earlier. In fact, over these 110 instances, o1-preview only manages 23.63%, and most of this accuracy comes from correctly solving problems which require fewer than 28 steps.
Why it matters here: o1-preview scored 97.8% on small problems but 23.63% on 6-20 block problems, and its correct answers on the larger set were mostly the shorter plans. That is good performance on small problems with a quick fall-off.
- Publisher
- arXiv (Valmeekam, Stechly, Kambhampati)
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv (Valmeekam, Stechly, Kambhampati),
while o1's performance is a quantum improvement on the benchmark, outpacing the competition, it is still far from saturating it.
Why it matters here: The authors also call o1 a large improvement over earlier models. This is fair context: 'ok on small problems' if anything undersells the 97.8%.
- Where
- Abstract
- Publisher
- arXiv (Valmeekam, Stechly, Kambhampati)
- 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, LLMs don’t do formal reasoning - and that is a HUGE problem
Performance is ok on small problems, but quickly tails off.
- Where
- paragraph 11
- Published
- 2024-10-11
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-llms-dont-do-formal-reasoning-and-4a617e-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-llms-dont-do-formal-reasoning-and-4a617e-a1.