Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The LLM Lobotomy?

learn.microsoft.com

Post reply on HN
Print view View original

11–20 of 62 posts

Re: The LLM Lobotomy?

#11

Post by chaos_emergent »

briga wrote: post #10

I have a theory: all these people reporting degrading model quality over time aren't actually seeing model quality deteriorate. What they are actually doing is discovering that these models aren't as powerful as they initially thought (ie. expanding their sample size for judging how good the model is). The probabilistic nature of LLM produces a lot of confused thinking about how good a model is, just because a model…

Yes exactly, my theory is that the novelty of a new generation of LLMs’ performances tends to cause an inflation in peoples’ perceptions of the model, with a reversion to a better calibrated expectation over time. If the developer reported numerical evaluations that drifted over time, I’d be more convinced of model change.

Re: The LLM Lobotomy?

#12

Post by colordrops »

briga wrote: post #10

I have a theory: all these people reporting degrading model quality over time aren't actually seeing model quality deteriorate. What they are actually doing is discovering that these models aren't as powerful as they initially thought (ie. expanding their sample size for judging how good the model is). The probabilistic nature of LLM produces a lot of confused thinking about how good a model is, just because a model…

Did any of you read the article? They have a test framework that objectively shows the model getting worse over time.

Re: The LLM Lobotomy?

#13

Post by yieldcrv »

briga wrote: post #10

I have a theory: all these people reporting degrading model quality over time aren't actually seeing model quality deteriorate. What they are actually doing is discovering that these models aren't as powerful as they initially thought (ie. expanding their sample size for judging how good the model is). The probabilistic nature of LLM produces a lot of confused thinking about how good a model is, just because a model…

fta: “I am glad I have proof of this with the test system”

I think they have receipts, but did not post them there

Re: The LLM Lobotomy?

#14

Post by transcriptase »

SubiculumCode wrote: post #7
ProjectArcturis wrote: post #3

I'm confused why this is addressed to Azure instead of OpenAI. Isn't Azure just offering a wrapper around chatGPT? That said, I would also love to see some examples or data, instead of just "it's getting worse".

I don't remember where I saw it, but I remember a claim that Azure hosted models performed poorer than those hosted by openAI.

Explains why the enterprise copilot ChatGPT wrapper that they shoehorn into every piece of office365 performs worse than a badly configured local LLM.

Re: The LLM Lobotomy?

#15

Post by colordrops »

esafak wrote: post #2

This was the perfect opportunity to share the evidence. I think undisclosed quantization is definitely a thing. We need benchmarks to be periodically re-evaluated to ward against this. Providers should keep timestamped models fixed, and assign modified versions a new timestamp, and price, if they want. The model with the "latest" tag could change over time, like a Docker image. Then we can make an informed decision o…

In addition to quantization, I suspect the additions they make continually to their hidden system prompt for legal, business, and other reasons slowly degrade responses over time as well.

Re: The LLM Lobotomy?

#16

Post by jonplackett »

colordrops wrote: post #15
esafak wrote: post #2

This was the perfect opportunity to share the evidence. I think undisclosed quantization is definitely a thing. We need benchmarks to be periodically re-evaluated to ward against this. Providers should keep timestamped models fixed, and assign modified versions a new timestamp, and price, if they want. The model with the "latest" tag could change over time, like a Docker image. Then we can make an informed decision o…

In addition to quantization, I suspect the additions they make continually to their hidden system prompt for legal, business, and other reasons slowly degrade responses over time as well.

This is quite similar to all the modifications intel had to do due to spectre - I bet those system prompts have grown exponentially.

Re: The LLM Lobotomy?

#17

Post by zzzeek »

I'm sure MSFT will offer this person some upgraded API tier that somewhat improves the issues, though not terrifically, for only ten times the price.

Re: The LLM Lobotomy?

#18

Post by jonplackett »

cush wrote: post #8

Is setting temperature to 0 even a valid way to measure LLM performance over time, all else equal?

It could be that performance on temp zero has declined but performance on a normal temp is the same or better.

I wonder if temp zero would be more influenced by changes to the system prompt too. I can imagine it making responses more brittle.

Re: The LLM Lobotomy?

#19

Post by Aurornis »

yieldcrv wrote: post #13
briga wrote: post #10

I have a theory: all these people reporting degrading model quality over time aren't actually seeing model quality deteriorate. What they are actually doing is discovering that these models aren't as powerful as they initially thought (ie. expanding their sample size for judging how good the model is). The probabilistic nature of LLM produces a lot of confused thinking about how good a model is, just because a model…

fta: “I am glad I have proof of this with the test system” I think they have receipts, but did not post them there

A lot of the claims I’ve seen have claimed to have proof, but details are never shared.

Even a simple graph of the output would be better than nothing, but instead it’s just an empty claim.

Re: The LLM Lobotomy?

#20

Post by zzzeek »

briga wrote: post #10

I have a theory: all these people reporting degrading model quality over time aren't actually seeing model quality deteriorate. What they are actually doing is discovering that these models aren't as powerful as they initially thought (ie. expanding their sample size for judging how good the model is). The probabilistic nature of LLM produces a lot of confused thinking about how good a model is, just because a model…

your theory does not hold up for this specific article as they carefully explained they are sending identical inputs into the model each time and observing progressively worse results with other variables unchanged. (though to be fair, others have noted they provided no replication details as to how they arrived at these results.)
Post reply on HN