Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by jhrmnn on Friday December 20, 2024 @02:20PM
(#42474020)
That’s for the low-compute configuration that doesn’t reach human-level performance (not far though)
[ Parent
| on HN ]
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by riku_iki on Friday December 20, 2024 @02:21PM
(#42474028)
I referred on high compute mode. They have table with breakdown here: https://arcprize.org/blog/oai-o3-pub-breakthrough
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by junipertea on Friday December 20, 2024 @02:30PM
(#42474129)
The table row with 6k figure refers to high efficiency, not high compute mode. From the blog post:
Note: OpenAI has requested that we not publish the high-compute costs. The amount of compute was roughly 172x the low-compute configuration.
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by gbnwl on Friday December 20, 2024 @02:30PM
(#42474137)
That's "efficiency" high, which actually means less compute. The 87.5% score using low efficiency (more compute) doesn't have cost listed.
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by bluecoconut on Friday December 20, 2024 @02:31PM
(#42474141)
they use some poor language.
"High Efficiency" is O3 Low "Low Efficiency" is O3 High
They left the "Low efficiency" (O3 High) values as `-` but you can infer them from the plot at the top.
Note the $20 and $17 per task aligns with the X-axis of the O3-low
Re:OpenAI O3 breakthrough high score on ARC-AGI-PUB
by EVa5I7bHFq9mnYK on Friday December 20, 2024 @02:27PM
(#42474105)
That's high EFFICIENCY. High efficiency = low compute.