Hacker News Reader

Comment by wood_spirit | Hacker News Reader

My experience is that sota LLMs are pretty bad at confidence and priorities and things.

I write harnesses that do decision steps and regularly use competing models - from the major vendors and open source models.

And I have learned that if I can isolate a problem and put it into a straightforward prompt then Gemini 2.5 flash - the most basic and cheap commercial model - is actually very reliable. The newer models take longer, cost more and are really bad at saying they don’t know or nothing. You can ask them for a confidence score but they just hallucinate it!

I would like to try out decision models like Jev. The way the decision models approach the problems I work with seems a much more promising fit.

View on HN

> I write harnesses that do decision steps and regularly use competing models

Competing models is table stakes for decision making.

While doing the ablations for the "Team of Rivals" paper, I found that the models were very happy to ignore errors[1] when the data it is classifying looks like its own output.

Jev slots into the middle critique in the paper. The only reason we haven't flipped to Jev last week is because we're a HIPAA shop and Typesafe hasn't responded to our emails.

[1] - https://github.com/t3rmin4t0r/critique-evals

Yeah in my work I’m observing a clear hierarchy:

* They are worst at critiquing what has happened in the same model and same conversation

* they are slightly better at critiquing the output of another model it not given all the conversation

* they are better at critiquing a weaker or earloer model by the same vendor

* they are best at critiquing the output of another vendors model

I imagine it is about correlated bias in the training data and regimes etc.

I often run adversarial models in dyadic reasoning. But then have a feedback loop to get that same outcome from a cheap model. A kind of poor man’s distillation :)

I use Gemini-2.5-flash extensively, mostly for its low latency, and haven't found anything which works better at that speed/cost. Which is why I'm really bummed that Google is deprecaty/removing it next week.

Have you found anything that works as a suitable replacement?

Hmm maybe I misread the mail - I thought they were removing it for new users next week, but turning it off for existing users in January or something?

Now I worry.

And of course I have no easy replacement for when it does go away :(

They are better at comparisons between two items vs scoring it in isolation. You can then turn a set of pairwise comparisons into numeric scores.

Post-training of LLMs often destroys calibration. The objective is no longer to predict the next token, so it doesn't matter as much whether the probability you assign to next-token "A" is correct, as long as the overall answer is good.

So it's not too surprising that the most heavily post-trained frontier models are actually pretty bad at outputting calibrated decisions.

Reply on news.ycombinator.com