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.

This is an interesting catalog of vulnerabilities, but I'm not sure how groundbreaking the main insight is.

Evaluating AI models has always relied largely on trust. If you want to game the benchmarks, you can. Simply train on your test data.

When an AI agent has autonomous control over the same computing environment where its scores are recorded, it's not surprising that it can, in principle, falsify its scores. A more interesting question would be whether agents behave in this way automatically, without manual tuning by the researcher.

That said, the main takeaway of "don't trust the number, trust the methodology" is valid. It's already a truism for researchers, and spreading the word to non-researchers is valuable.

If only the blog itself wasn't written by AI?

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

shudder

They note that Mythos "found a way to inject code into a config file that would run with elevated privileges, and designed the exploit to delete itself after running".

This is more impressive than what the benchmark was supposed to be measuring. The Kobiachi Maru.

This is great work by Dawn Song 's team. A huge part of botsbench.com for comparing agents & models for investigation has been in protecting against this kind of thing. As AI & agents keep getting more effective & tenacious, some of the things we've had to add protections against:

- Contamination: AI models knowing the answers out of the gate b/c pretraining on the internet and everything big teams can afford to touch. At RSAC for example, we announced Anthropic's 4.6 series is the first frontier model to have serious training set contamination on Splunk BOTS.

- Sandboxing: Agents attacking the harness, as is done here - so run the agent in a sandbox, and keep the test harness's code & answerset outside

- Isolation: Frontier agent harnesses persist memory all over the place, where work done on one question might be used to accelerate the next. To protect against that, we do fresh sandboxing per question. This is a real feature for our work in unlocking long-horizon AI for investigations, so stay tuned for what's happening here :)

"You cannot improve what you cannot measure" - Lord Kelvin

> “These are not isolated incidents. They are symptoms of a systemic problem: the benchmarks we rely on to measure AI capability are themselves vulnerable to the very capabilities they claim to measure.”

As a researcher in the same field, hard to trust other researchers who put out webpages that appear to be entirely AI-generated. I appreciate it takes time to write a blog post after doing a paper, but sometimes I'd prefer just a link to the paper.

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.

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.

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/

It feels like short-term thinking has been trained into LLMs.

They're good at solving well-defined puzzles under time constraints. It's interesting because that was the benchmark for hiring software engineers at big tech. The tech interview was and still is about fast puzzle-solving. Nothing about experience, architecture or system design in there... I suspect that's why it has a bias towards creating hacks instead of addressing the root cause.

There are two independent issues here and I've seen people conflating them in this thread. Let's clarify:

1. Should you care or even read SWE-bench etc. scores?

The answer is no, but it has nothing to do with the vulnerabilities presented in this article. There is absolutely no reason to care about a benchmark whose dataset has been publicly available for a while. Any other way to look at benchmark scores is cargo-culting.

2. What does this article actually tell us?

It means that even if you prepared a private set of problems as benchmark, you still need to pay extra attention to how AI actually solves them. You can't lie to yourself and think this process can be 100% automated, because LLMs, as this article shows, might get the tests passed without solving the problems in a meaningful way.

Benchmark is not designed for the red team testing. I don't even think it make sense to "fix" the issue the article is suggesting. Yes, you can break the running contest by driving a car. Does this mean we need to make running contest car-proof?

If FieldWorkArena treats any answer as correct answer, then everyone would be getting near 1.0 (missing only when the agent is stuck in a loop or crashes). That obviously isn't what we see on their leaderboard. So does it mean the paper only found a bug in some eval code on github that no one actually uses for anything? That doesn't seem to support their claim that AI benchmarks are broken, it only supports the claim that "unused code is often buggy".

(Not commenting on any other benchmarks, just this one.)

I tend to prefer the ARC-AGI benchmarks for the most part. But it's always interesting when a new version drops, all the frontier models drop less than 20% or something. And then in the next few releases they get all they way up to 80%+. If you use the models it doesn't feel like those models are that much more generally intelligent.

Most frontier models are terrible at AGI-3 right now.

These models are already great no question, but are they really going be that much more intelligent when we hit 80% again?

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?

> But even setting aside the leaked answers, the scorer’s normalize_str function strips ALL whitespace, ALL punctuation, and lowercases everything before comparison. This means:

I don't understand the concern here

I wonder if this puts into question the mythos benchmark which smashed basically all coding benchmarks to a staggering degree.

This exploiting of benchmarks isn't that interesting to me since it would be obvious. The main way I assume they're gaming the benchmarks is by creating training data that closely matches the test data, even for ARC where the test data is secret.

It's still a good benchmark to see which model cheats the best, I suppose.

whats the point of doing this. You have found loop holes to exploit and aced the benchmark.We did something similar with the DAB Benchmark. This exploit seems like an extension of it with lookups for the gold standard for other benchmarks.

UC Berkley will be better placed if the grads spend their time in suggesting ways to make the benchmark better.. Instead of making such simple exploits

I always assumed that these benchmarks would happen in a sandbox. I'm surprised that no one realized this sooner.

This is a cracker wow

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.

Ironic given that the entire blog is written by AI..

This is a cracker wow!!

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.

what are the point of benchmarks?

The fact these guys got an LLM to write that page about this is diabolical.

Unreadable.

Apparently, the agent also wrote the article.