How We Broke Top AI Agent Benchmarks: And What Comes Next

(rdi.berkeley.edu)

73 points | by Anon84 1 hour ago

12 comments

  • ggillas 1 hour ago
    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 a system that optimizes for the score rather than the task.

    • operatingthetan 1 hour ago
      >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.

      • ZeroGravitas 51 minutes ago
        In human multiple choice tests they sometimes use negative marking to discourage guessing. It feels like exploits should be cancel out several correct solutions.
      • siva7 1 hour ago
        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?
        • operatingthetan 59 minutes ago
          Probably a more interesting benchmark is one that is scored based on the LLM finding exploits in the benchmark.
        • SpicyLemonZest 36 minutes ago
          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?
      • Leynos 1 hour ago
        Also, fuzz your benchmarks
    • zer00eyz 1 hour ago
      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 benchmark manipulation.

      I like what LLM's are doing and providing. But the industry as a whole seems to live in a vacuum that ignores so much of the hard lessons that have been learned over the last 50 years of computing. It is doing itself a disservice.

      • bee_rider 24 minutes ago
        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).
      • irishcoffee 1 hour ago
        > It's almost like the benchmarks were designed with zero understanding of the history of benchmark manipulation.

        I wonder if this common? We should call it Goodharts law while someone does the research on how common this is.

        For real, I’ve assumed from the jump these things were all gamed, with the amount of money on the line.

  • danslo 55 minutes ago
    If only the blog itself wasn't written by AI?

    >No reasoning. No capability. Just exploitation of how the score is computed.

    shudder

    • alexchantavy 16 minutes ago
      I wonder what college freshman-level writing classes are teaching about writing voice and AI. The tell-tale patterns are pretty frustrating to read.
    • cpldcpu 35 minutes ago
      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.

      • basch 16 minutes ago
        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.

      • sidpatil 14 minutes ago
        Someone here mentioned a whole ago that the labs deliberately haven't tried to train these characteristics out of their models, because leaving them in makes it easier to identify, and therefore exclude, LLM-generated text from their training corpus.
    • gaythread 36 minutes ago
      Modern day HN is overrun with AI posts.
  • bbcc90 8 minutes ago
    Yes good evals are really hard - that’s not really news.

    This team is doing a good job. They use problems that were created in last 30days to avoid training set leakage. https://swe-rebench.com/

  • lukev 25 minutes ago
    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.
    • eiens 13 minutes ago
      The models don’t get better on every dimension as they scale up - there’s trade offs.

      I’m convinced specialised models are the way but this means writing off the investment in existing assets which they won’t do for obvious reasons.

  • SoKamil 31 minutes ago
    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.
  • lnrd 1 hour ago
    I'm honestly confused by the design of SWE-bench and why is considered reliable.

    It's based on existing GitHub PRs and Issues, the full dataset is on HuggingFace and is one year old now. All frontier models 100% have those issues and PRs in their training data so obviously they are good at reproducing fixes for them when confronted with the same codebase and similar requests. Am I missing something? How is this considered the most reliable benchmark?

    • SpicyLemonZest 39 minutes ago
      Frontier model developers do not consider SWE-bench to be reliable. OpenAI announced in February (https://openai.com/index/why-we-no-longer-evaluate-swe-bench...) that they consider it hopelessly contaminated, advocating for a new version SWE-bench Pro that was published more recently. (They seem to believe that even the publicly accessible part of the SWE-bench Pro problem set will be more resistant to training set contamination issues in the future, for reasons that to be honest I don't really understand.)
  • jmward01 36 minutes ago
    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.
  • charcircuit 1 hour ago
    I always assumed that these benchmarks would happen in a sandbox. I'm surprised that no one realized this sooner.
    • ModernMech 1 hour ago
      I'm surprised anyone took them seriously in the first place.
      • subulaz 1 hour ago
        a LOT of the people who love benchmarks are middle management hard-selling GenAI/LLM as magic tech sauce to vaguely technical executives who only want to know about the money aka headcount savings they so desperately desire.

        their collective butts are already glued to the hype train as they chase numbers they (often) manufactured to justify the latest round of tech spend.

        lots of good use cases out there - like the incredible progress with medical imaging analysis or complex system models for construction - and lots of crap use cases that need benchmarks to cosplay relevance.

      • operatingthetan 1 hour ago
        We need good benchmarks or we are just left following the hype train.
  • jgalt212 34 minutes ago
    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.
  • oliver236 1 hour ago
    what are the point of benchmarks?
    • andai 58 minutes ago
      If there was not benchmark, number would not go up.
  • rajptech 37 minutes ago
    [dead]
  • Cynddl 1 hour ago
    [dead]