Researchers asked frontier models to play chess. Have a look at the MAR rates in Table 3. When not explicitly told which moves were legal, no model identified legal moves at a rate better than 80%. Many asked for more illegal moves than legal moves. And even when explicitly told which moves were legal, the models continued to ask for illegal moves. With illegal asks discarded, none of the bots could beat a chess model calibrated to 1100 ELO.
The author of the originating post says that "current frontier models need laborious oversight and guardrails on even the simplest tasks", and he's absolutely correct.
> Gemini 2.5 Pro, O3, Claude Sonnet 3.7 and ChatGPT 4.1
The gap in capabilities between those models which they tested, and actual current frontier ones is enormous. I would not trust that any conclusions they made are applicable.
Is it? I am on the fence on this, but it does seem like there are some qualitative improvements between the models.
Not related to your post, but a fact I keep mulling over. The fact I don't trust the current crop of LLM's enough and I consider LLM's as a tech will hit a ceiling pretty hard, it doesn't mean parallel improvement curves won't spring up out of other research that will lead to much higher capabilities than currently.
> but it does seem like there are some qualitative improvements between the models.
It could easily seem that way, I think, in a "quantity has a quality of its own" kind of way. When you can come to the same conclusion faster, that lets you iterate more; and sometimes when you iterate you find more things.
carodgers · · focus · HN ↗
<a href="https://arxiv.org/html/2509.24239v4" rel="nofollow">https://arxiv.org/html/2509.24239v4
Researchers asked frontier models to play chess. Have a look at the MAR rates in Table 3. When not explicitly told which moves were legal, no model identified legal moves at a rate better than 80%. Many asked for more illegal moves than legal moves. And even when explicitly told which moves were legal, the models continued to ask for illegal moves. With illegal asks discarded, none of the bots could beat a chess model calibrated to 1100 ELO.
The author of the originating post says that "current frontier models need laborious oversight and guardrails on even the simplest tasks", and he's absolutely correct.
joefourier · · focus · HN ↗
> Gemini 2.5 Pro, O3, Claude Sonnet 3.7 and ChatGPT 4.1
The gap in capabilities between those models which they tested, and actual current frontier ones is enormous. I would not trust that any conclusions they made are applicable.
dgb23 · · focus · HN ↗
RealityVoid · · focus · HN ↗
Not related to your post, but a fact I keep mulling over. The fact I don't trust the current crop of LLM's enough and I consider LLM's as a tech will hit a ceiling pretty hard, it doesn't mean parallel improvement curves won't spring up out of other research that will lead to much higher capabilities than currently.
zahlman · · focus · HN ↗
It could easily seem that way, I think, in a "quantity has a quality of its own" kind of way. When you can come to the same conclusion faster, that lets you iterate more; and sometimes when you iterate you find more things.