Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by siva7 on Saturday April 11, 2026 @04:07PM
(#47733557)
Could it really be that not only we vibeslop all apps nowadays but also don't care to even check how ai solved a benchmark it claimed solved?
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by operatingthetan on Saturday April 11, 2026 @04:16PM
(#47733637)
Probably a more interesting benchmark is one that is scored based on the LLM finding exploits in the benchmark.
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by SpicyLemonZest on Saturday April 11, 2026 @04:39PM
(#47733827)
Frontier model developers try to check for memorization. But until AI interpretability is a fully solved problem, how can you really know whether it actually didn't memorize or your memorization check wasn't right?
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by retinaros on Saturday April 11, 2026 @06:39PM
(#47734624)
Every ai labs train on the test set. That is a big part of why we see benchmark climbing from 1% to 30% after a few models iterations
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by latentsea on Saturday April 11, 2026 @11:57PM
(#47736026)
Models themselves definitely aren't getting better.
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by stingraycharles on Sunday April 12, 2026 @01:09AM
(#47736305)
This is already well known, all these AI benchmarks use a different model to judge whether or not the solution was correct.
It’s… remarkably poor, and as demonstrated in the paper, easily gamed. Worst yet, these benchmarks teach AIs to be very short-sighted and hyper-focused on completing the task, rather than figuring out the best solution.
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by Leynos on Saturday April 11, 2026 @04:02PM
(#47733526)
Also, fuzz your benchmarks
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by ZeroGravitas on Saturday April 11, 2026 @04:24PM
(#47733711)
In human multiple choice tests they sometimes use negative marking to discourage guessing. It feels like exploits should cancel out several correct solutions.
Re:How We Broke Top AI Agent Benchmarks: And What Comes Next
by lambda on Saturday April 11, 2026 @05:29PM
(#47734176)
Unfortunately, very few LLM benchmarks do this. LLMs get such high scores on many benchmarks because there's no difference between answering "I don't know" as giving a made up answer, and made up answers can improve the score some of the time, so by chasing higher benchmark numbers on these kinds of benchmarks, the labs are prioritizing guessing over accuracy.
The Artificial Analysis Omniscience benchmark does penalize guessing, so it actually helps you determine which LLMs are likely to just guess rather than telling you they don't know. Only a very few of the frontier models actually score higher than 0 on this, where 0 means that it's equally likely to return a correct answer as it is to return a hallucination on factual questions.
Yeah the path forward is simple: check if the solutions actually contain solutions. If they contain exploits then that entire result is discarded.