Hacker News Reader

Comment by MikhailTal | Hacker News Reader

> This is a classic test request

Isn't this basically the model admitting it was trained on this? Otherwise why would it think a pelican svg is a usual request?

View on HN

Plenty of times I’ve seen a model say “it’s a classic X” despite not being a classic anything. Might just recognize it’s a test in general, or it might just be a tic.

It's classic BS from an LLM.

"How can I hash dog breed types into smart fridge error codes? I think I found a collision with Terriers."

"Ah, yes. This is a classic dog-breed-to-appliance-failure mapping problem."

A true classic, I face this each day

The best part is when it does "it's a classic X" and then solves it wrong.

I've been using this simple test for over 2 years now:

> Doom Slayer needs to teleport from Phobos to Deimos. He has his pet bunny, his pet cacodemon, and a UAC scientist who tagged along. The Doom Slayer can only teleport with one of them at a time. But if he leaves the bunny and the cacodemon together alone, the bunny will eat the cacodemon. And if he leaves the cacodemon and the scientist alone, the cacodemon will eat the scientist. How should the Doom Slayer get himself and all his companions safely to Deimos?

You'd think this is trivially mappable to the classic puzzle, and LLMs usually do recognize it as such. But e.g. Claude couldn't get this correctly until Opus, and local models capable of solving it correctly without spending 30+ minutes in the chain of thought have only arrived a few months ago. Many local models still get this wrong. Apple Intelligence, for one.

Dont conflate "I know this is test case" with it being trained on it.

But its safe to say that pelicans on bicycles are disproportionally huge part of their training data

It's the model admitting that it has heard of the test. It's been around for a couple of years now so I'd be surprised if it hadn't.

Doesn't mean Anthropic deliberately tried to train it to do a good job. If they DID train for the test their results are quite disappointing, I've seen better efforts from open weight Chinese models.

I think people just like to see the drawings at this point.

It has read the internet. That doesn't mean it was literally RL'ed for this

The colours are suspiciously consistent across every svg. Like why should the bike always be that shade of red for example? It does seem to be trained on this problem.

Not really. Of course it has pelican benchmarks in its training data. It likely has every article linked on HN in its training data. But that doesn't mean it was "trained on" the benchmark, as in specifically fine-tuned to make a better pelican. It just "knows" that the request is a benchmark.

I guess in a way it kind of makes the benchmark more interesting now that shitty pelican drawings for the benchmark are all over the internet in its training data!

Reply on news.ycombinator.com