Exploiting the most prominent AI agent benchmarks • Hacker News Board

Re: Exploiting the most prominent AI agent benchmarks

#21

Post by SpicyLemonZest » Sat, Apr 11, 2026, 8:39 PM UTC

siva7 wrote: post #11

operatingthetan wrote: post #5

Earlier quoted context omitted.

>hopefully changes the way benchmarking is done. Yeah the path forward is simple: check if the solutions actually contain solutions. If they contain exploits then that entire result is discarded.

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?

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: Exploiting the most prominent AI agent benchmarks

#22

Post by gaythread » Sat, Apr 11, 2026, 8:39 PM UTC

danslo wrote: post #17

If only the blog itself wasn't written by AI? >No reasoning. No capability. Just exploitation of how the score is computed. shudder

[flagged]

Re: Exploiting the most prominent AI agent benchmarks

#23

Post by jmward01 » Sat, Apr 11, 2026, 8:39 PM UTC

Not really on the topic, but I have wondered if we need a different type of test to help find model architecture potential. Standardized training sets followed by testing to see the potential curves of a model. train on x, test, add y, test, add z, test. At each increment you see how well the model is absorbing the information and extrapolate how well that architecture may do if more fully trained.

Re: Exploiting the most prominent AI agent benchmarks

#24

Post by cpldcpu » Sat, Apr 11, 2026, 8:39 PM UTC

danslo wrote: post #17

If only the blog itself wasn't written by AI? >No reasoning. No capability. Just exploitation of how the score is computed. shudder

Yes, marks of AI all over the place. Also the SVGs.

>No solution written, 100% score.

Its weird. Turns out that hardest problem for LLMs to really tackle is long-form text.

Re: Exploiting the most prominent AI agent benchmarks

#25

Post by jgalt212 » Sat, Apr 11, 2026, 8:41 PM UTC

The real question is how to close to VW and Deiselgate are these offenses? And what exposure do these companies have? I would assume securities fraud, if only because Matt Levine says everything is securities fraud.

Re: Exploiting the most prominent AI agent benchmarks

#26

Post by SoKamil » Sat, Apr 11, 2026, 8:43 PM UTC

The more research on this topic is created, the more knowledge how to game them will be stored in future training data. And since it comes from university, it is ranked higher in data corpus. It sounds like a self fulfilling prophecy.

Re: Exploiting the most prominent AI agent benchmarks

#27

Post by abirch » Sat, Apr 11, 2026, 8:45 PM UTC

SoKamil wrote: post #26

The more research on this topic is created, the more knowledge how to game them will be stored in future training data. And since it comes from university, it is ranked higher in data corpus. It sounds like a self fulfilling prophecy.

Damned old Goodhart's Law: "When a measure becomes a target, it ceases to be a good measure".

https://en.wikipedia.org/wiki/Goodhart%27s_law

Re: Exploiting the most prominent AI agent benchmarks

#28

Post by lukev » Sat, Apr 11, 2026, 8:50 PM UTC

I think we should all consider the possibility that part of the reason Anthropic hasn't immediately released Mythos is that it would be slightly disappointing relative to the benchmark scores.

Re: Exploiting the most prominent AI agent benchmarks

#29

Post by bee_rider » Sat, Apr 11, 2026, 8:51 PM UTC

zer00eyz wrote: post #6

ggillas wrote: post #2

This is a phenomenal paper on exploits and hopefully changes the way benchmarking is done. From the paper: We achieved near-perfect scores on all of them without solving a single task. The exploits range from the embarrassingly simple (sending {} to FieldWorkArena) to the technically involved (trojanizing binary wrappers in Terminal-Bench), but they all share a common thread: the evaluation was not designed to resist…

2024: Industry group invalidates 2,600 official Intel CPU benchmarks — SPEC says the company's compiler used unfair optimizations to boost performance https://www.tomshardware.com/pc-components/cpus/spec-invalid... 2003: Nvidia accused of cheating in 3DMark 03 https://www.gamespot.com/articles/nvidia-accused-of-cheating... It's almost like the benchmarks were designed with zero understanding of the history of benchma…

What was the cheat in the 2024 Intel situation? The TomsHardware article and the Phoronix article they linked were quite vague. (Not to say I have any doubts, just curious, hadn’t heard of this one).

Re: Exploiting the most prominent AI agent benchmarks

#30

Post by basch » Sat, Apr 11, 2026, 8:59 PM UTC

cpldcpu wrote: post #24

danslo wrote: post #17

If only the blog itself wasn't written by AI? >No reasoning. No capability. Just exploitation of how the score is computed. shudder

Yes, marks of AI all over the place. Also the SVGs. >No solution written, 100% score. Its weird. Turns out that hardest problem for LLMs to really tackle is long-form text.

Maybe in one shot.

In theory I would expect them to be able to ingest the corpus of the new yorker and turn it into a template with sub-templates, and then be able to rehydrate those templates.

The harder part seems to be synthesizing new connection from two adjacent ideas. They like to take x and y and create x+y instead of x+y+z.