Re:MAI-Code-1-Flash
by OsrsNeedsf2P on Tuesday June 02, 2026 @02:55PM
(#48374585)
So it's trained on the SWE Bench Pro evalset
[ Parent
| on HN ]
Re:MAI-Code-1-Flash
by lemonish97 on Tuesday June 02, 2026 @02:56PM
(#48374609)
What is your evidence for this claim?
They say hill climbing
https://microsoft.ai/news/building-a-hillclimbing-machine-la...
Unless they specifically clarify that the testing and training benchmarks are completely separate, we have to assume they test on the same 'hill' the model climbs.
Re:MAI-Code-1-Flash
by artemisart on Tuesday June 02, 2026 @04:47PM
(#48376049)
Hill climbing doesn't mean much but absolutely doesn't imply they cheat on benchmarks.
They have more details here https://microsoft.ai/news/introducing-mai-thinking-1/ it seems to be "RL on everything".
That's not accurate. Take a look at the paper to see what it is trained on! And specifically decontamination is called out in A.4
https://microsoft.ai/wp-content/uploads/2026/06/main_2026060...