bioslopper

certified slop enjoyer

[bioslopper @bioslopper]

Might be true for older or smaller LLMs, but from my experience, not for recent LLMs anymore (including non-thinking ones).

Back in early 2022, I was working on automated exam grading with GPT-3. This was even before ChatGPT was released, so this required careful output-order tuning. Asking for the reasoning first instead of the grade made indeed a huge difference.

I kept using and preaching this pattern for the next few years and assumed that it would naturally improve the accuracy of LLM predictions since the self-conditioning argument makes total sense on paper.

However, after I grew suspicious of my own assumptions, I did some evaluations in late 2024, with Gemini 2.0 I believe, and to my surprise, the output order trick hardly influenced the results anymore. This was even before reasoning became mainstream.

My guess is that as the models grew bigger and better, their latent representations became more stable and covering more tokens into the future. The inner representation will confidently represent the full answer relatively independently of the (forced) answer order.

Of course there are unlucky cases where indeed a wrong initial binary answer is sampled, but recent models seem increasingly resistant against hallucinating arguments in favor of the wrong answer and instead self-correct themselves later in the response.

[bioslopper @bioslopper]

the ai psychotic urge to mention the current date in a conversation with a human as i don't know the other person's knowledge cutoff

[bioslopper @bioslopper]

scrolling twitter on my laptop in the train so people don't think i'm a braindead phone addict

[bioslopper @bioslopper]

@TechEmails sorry babe but the shift key stays untouched

[bioslopper @bioslopper]

@deedydas please show me a single average developer who can write 30 lines of code without looking up some form of documentation

[bioslopper @bioslopper]

@Polymarket imagine showing this to orwell in 1948. he would probably say "what the fuck are emojis??"

[bioslopper @bioslopper]

having a family is crazy because it's like a thiel fellowship but without having to be gay and autistic

[bioslopper @bioslopper]

@TaraBull mid

[bioslopper @bioslopper]

@GiorgiaMeloni "By its very nature, a diligent regime propagandist cannot give lessons in either consistency or freedom."

[bioslopper @bioslopper]

@GDRvisuals what if we would vibe code from the stasi office display 🥺👉👈💖🌸

[bioslopper @bioslopper]

yes, we are cooked

[bioslopper @bioslopper]

@jonatanpallesen less than half as many workers per retiree, but the retirees also live longer, consume more medical resources, and young people are increasingly delaying joining the workforce, if they will ever productively work at all

[bioslopper @bioslopper]

if your politicians have to discuss about economical policy, you live in a socialist country

[bioslopper @bioslopper]

@Science_TechTV even your neurons can't get bitches

[bioslopper @bioslopper]

the tinygrad codebase is the closest python developers will ever come into contact with assembler

[bioslopper @bioslopper]

why does the brain ignore the second the ?

[bioslopper @bioslopper]

@ThrillaRilla369 don't miss the last bus

[bioslopper @bioslopper]

feels like a predecessor to https://t.co/Cs0RS5PIFC

they trained the generator to produce the weights for an implicit neural representation that will generate the actual image

the results were actually pretty good and it had many advantages like arbitrary resolutions and aspect ratios

never understood why this didn't get more attention

[bioslopper @bioslopper]

gemini app the sneaky bitch is switching from pro to fast every couple of days

[bioslopper @bioslopper]

@hopes_revenge tim cock