AIAI for scienceStory 04
High science scores still hide unfinished research
What changedPhysClosureBench tests fifty expert-verified physics investigations and scores both local accomplishments and complete research closure. The top model averaged 6.29 of 10 but reached full closure in only 32.7 percent of attempts; 36.2 percent of attempts scoring at least 8 still remained incomplete.
The useful part
Why it matters
The top model averaged 6.29 of 10 but reached full closure in only 32.7 percent of attempts; 36.2 percent of attempts scoring at least 8 still remained incomplete.
Keep in mind
Good to know
The benchmark covers a specialized set of physics tasks.
Evidence