Skip to content(if available)orjump to list(if available)

Adaptive LLM routing under budget constraints

pbd

GPT-4 at $24.7 per million tokens vs Mixtral at $0.24 - that's a 100x cost difference! Even if routing gets it wrong 20% of the time, the economics still work. But the real question is how you measure 'performance' - user satisfaction doesn't always correlate with technical metrics.

simpaticoder

PPT (price-per-token) is insufficient to compute cost. You will also need to know an average tokens-per-interaction (TPI). They multiply to give you a cost estimate. A .01x PPT is wiped out by 100x TPI.

FINDarkside

It's trivial to get better score than GPT-4 with 1% of the cost by using my propertiary routing algorithm that routes all requests to Gemini 2.5 Flash. It's called GASP (Gemini Always, Save Pennies)

nutjob2

Does anyone working in an individual capacity actually end up paying for Gemini (Flash or Pro)? Or does Google boil you like a frog and you end up subscribing?

mkoubaa

> How you measure 'performance'

I heard the best way is through valuations

Keyframe

number of complaints / million tokens?

pqtyw

> GPT-4 at $24.7 per million tokens

While technically true why would you want to use it when OpenAI itself provides a bunch of many times cheaper and better models?

KTibow

RouterBench is from March 2024.

spoaceman7777

Incredible that they are using contextual bandits, and named it: Preference-prior Informed Linucb fOr adaptive rouTing (PILOT)

Rather than the much more obvious: Preference-prior Informed Linucb For Adaptive Routing (PILFAR)

QuadmasterXLII

The framing in the headline is interesting. As far as I recall, spending 4x more compute on a model to improve performance by 7% is the move that has worked over and over again up to this point. 101 % of GPT-4 performance (potentially at any cost) is what I would expect an improved routing algorithm to achieve.

dang

(The submitted title was "93% of GPT-4 performance at 1/4 cost: LLM routing with weak bandit feedback")

fny

Is there a reason human preference data is even needed? Don't LLMs already have a strong enough notion of question complexity to build a dataset for routing?

delichon

> a strong enough notion of question complexity

Aka Wisdom. No, LLMs don't have that. Me neither, I usually have to step in the rabbit holes in order to detect them.

fny

"Do you think you need to do high/medium/low amount of thinking to answer X?" seems well within an LLMs wheelhouse if the goal is to build an optimized routing engine.

nutjob2

How do you think that an LLM could come by that information? Do you think that LLM vendors are logging performance and feeding that back into the model or some other mechanism?

jibal

LLMs don't have notions ... they are pattern matchers against a vast database of human text.

mhh__

Please do a SELECT * from this database

andrewflnr

Is this really the frontier of LLM research? I guess we really aren't getting AGI any time soon, then. It makes me a little less worried about the future, honestly.

Edit: I never actually expected AGI from LLMs. That was snark. I just think it's notable that the fundamental gains in LLM performance seem to have dried up.

kenjackson

First, I don't think we will ever get to AGI. Not because we won't see huge advances still, but AGI is a moving ambiguous target that we won't get consensus on.

But why does this paper impact your thinking on it? It is about budget and recognizing that different LLMs have different cost structures. It's not really an attempt to improve LLM performance measured absolutely.

_heimdall

So you don't expect AGI to be possible ever? Or is your concern mainly with the wildly different definitions people use for it and that we'll continue moving goal posts rather than agree we got there?

nutjob2

There's no concrete evidence AGI is possible mostly because it has no concrete definition.

It's mostly hand waving, hype and credulity, and unproven claims of scalability right now.

You can't move the goal posts because they don't exist.

ctoth

Is a random paper from Fujitsu Research claiming to be the frontier of anything?

andrewflnr

Not just this paper, but model working shenanigans also seem to have been a big part of GPT-5, which certainly claims to be frontier work.

srekhi

I'm not following this either. You'd think this would be frontier back in 2023

jibal

LLMs are not on the road to AGI, but there are plenty of dangers associated with them nonetheless.

andrewflnr

Agreed, broadly. I never really thought they were, but seeing people work on stuff like this instead of even trying to improve the architecture really makes it obvious.

nicce

Just 2 days ago Gemini 2.5 Pro tried to recommend me tax evasion based on non-existing laws and court decisions. The model was so charming and convincing, that even after I brought all the logic flaws and said that this is plain wrong, I started to doubt myself, because it is so good at pleasing, arguing and using words.

And most would have accept the recommendation because the model sold it as less common tactic, while sounding very logical.

nutjob2

Or you could understand the tool you are using and be skeptical of any of its output.

So many people just want to believe, instead of the reality of LLMs being quite unreliable.

Personally it's usually fairly obvious to me when LLMs are bullshitting probably because I have lots of experience detecting it in humans.

roywiggins

> even after I brought all the logic flaws and said that this is plain wrong

Once you've started to argue with an LLM you're already barking up the wrong tree. Maybe you're right, maybe not, but there's no point in arguing it out with an LLM.

yahoozoo

That and LLMs are seemingly plateauing. Earlier this year, it seemed like the big companies were releasing noticeable improvements every other week. People would joke a few weeks is “an eternity” in AI…so what time span are we looking at now?

muldvarp

There have been very large improvements in code generation in the last 6 months. A few weeks without improvement are not necessarily a plateau.

andrewflnr

That's just the thing. There don't seem to have been any breakthroughs in model performance or architecture, so it seems like we're back to picking up marginal reductions in cost to make any progress.

yieldcrv

just because it’s on arxiv doesn’t mean anything

arxiv is essentially a blog under an academic format, popular amongst asian and south asian academic communities

currently you can launder reputation with it, just like “white papers” in the crypto world allowed for capital for some time

this ability will diminish as more people catch on

guluarte

I'm starting to think that there will not be an 'AGI' moment, we will simply slowly build smarter machines over time until we realize there is 'AGI'. It would be like video calls in the '90s everybody wanted them, now everybody hates them, lmao.

nutjob2

Or we'll realize that human intelligence and machine intelligence is apple and oranges.