Remix.run Logo
▲ ranyume 2 hours ago

>Run decision models locally.

>example is a text classification task instead of a decision

▲hbrn an hour ago | parent | next [-]

"Decision model" is just marketing jargon.

decision model = classifier

system one model = small non-reasoning LLM

noul = boolean

confidence = f(probabilities)

It's sad to see how gullible engineers are today.

▲OgAstorga an hour ago | parent | prev | next [-]

text classification is equivalente to decision. This is exactly the same thing Jev does.

▲ranyume an hour ago | parent | next [-]

If it has four legs, a tail and barks why not call it a dog?

▲seemaze 6 minutes ago | parent | next [-]

This dog only barks when given biscuits

▲gchamonlive an hour ago | parent | prev [-]

Because this specific dog only barks in structured text

▲ricardobeat an hour ago | parent | prev | next [-]

It is not. In a benchmark with actual decisions - navigation, traffic, waypoints - laya does slightly better than a small classifier, with very low correlation to state changes.

▲cobanov an hour ago | parent [-]

[dead]

▲abirch an hour ago | parent | prev [-]

Jev does it more efficiently because it doesn't use an LLM https://typesafe.ai/blog/introducing-system-one-models-and-j...

▲rockinghigh 14 minutes ago | parent [-]

Their marketing language is misleading. They must still use some transformer language model backbone to encode the text input (BERT or decoder-only LLM). The biggest difference is the output, instead of auto-regressively generating tokens, they produce probabilities over a bounded set of decisions (more flexible classification).

▲kgeist 6 minutes ago | parent [-]

Stock LLMs already internally produce probabilities before selecting the next token, they're just currently predominantly post-trained for chat/tool calls, not for selecting options A/B/C in the first token.

▲cobanov an hour ago | parent | prev [-]

Fair point, that example is basically classification. I'll change it to something that looks more like a real decision.