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

The case for the return of fine-tuning

welovesota.com

Post reply on HN
Print view View original

11–20 of 85 posts

Re: The case for the return of fine-tuning

#11

Post by CuriouslyC »

HarHarVeryFunny wrote: post #7
CuriouslyC wrote: post #4

Fine tuning by pretraining over a RL tuned model is dumb AF. RL task tuning works quite well.

You may have no choice in how the model you are fine tuning was trained, and may have no interest in verticals it was RL tuned for. In any case, platforms like tinker.ai support both SFT and RL.

Why would you choose a model where the trained in priors don't match your use case? Also, keep in mind that RL'd in behavior includes things like reasoning and how to answer questions correctly, so you're literally taking smart models and making them dumber by doing SFT. To top it off, SFT only produces really good results when you have traces that closely model the actual behavior you're trying to get the model to display. If you're just trying to fine tune in a knowledge base, a well tuned RAG setup + better prompts win every time.

Re: The case for the return of fine-tuning

#12

Post by imcritic »

CuriouslyC wrote: post #11
HarHarVeryFunny wrote: post #7

Earlier quoted context omitted.

You may have no choice in how the model you are fine tuning was trained, and may have no interest in verticals it was RL tuned for. In any case, platforms like tinker.ai support both SFT and RL.

Why would you choose a model where the trained in priors don't match your use case? Also, keep in mind that RL'd in behavior includes things like reasoning and how to answer questions correctly, so you're literally taking smart models and making them dumber by doing SFT. To top it off, SFT only produces really good results when you have traces that closely model the actual behavior you're trying to get the model to d…

Because you need a solution for your problem and the available tools are what they are and nothing else and you don't have enough resources to train your own model.

Re: The case for the return of fine-tuning

#14

Post by soVeryTired »

meander_water wrote: post #10

A couple of examples I have seen recently which makes me agree with OP: - PaddleOCR, a 0.9B model that reaches SOTA accuracy across text, tables, formulas, charts & handwriting. [0] - A 3B and 8B model which performs HTML to json extraction at GPT-5 level accuracy at 40-80x less cost, and faster inference. [1] I think it makes sense to fine tune when you're optimizing for a specific task. [0] https://huggingface.co/p…

Have you used PaddleOCR? I'm surprised they're claiming SOTA without comparing against Amazon Textract or Azure doc intelligence (LayoutLM v3 under the hood, as far as I know).

I've played around with doc recognition quite a bit, and as far as I can tell those two are best-in-class.

Re: The case for the return of fine-tuning

#15

Post by libraryofbabel »

matusp wrote: post #6

Fine-tuning is a good technique to have in a toolbox, but in reality, it is feasible only in some use cases. On one hand, many NLP tasks are already easy enough for LLMs to have near perfect accuracy and fine tuning is not needed. On the other hand, really complex tasks are really difficult to fine-tune and clevem data collection might be pretty expensive. Fine-tuning can help with the use cases somewhere in the midd…

What would you say is an example of one of those “middle” tasks it can help with?

Re: The case for the return of fine-tuning

#16

Post by CaptainOfCoit »

libraryofbabel wrote: post #15
matusp wrote: post #6

Fine-tuning is a good technique to have in a toolbox, but in reality, it is feasible only in some use cases. On one hand, many NLP tasks are already easy enough for LLMs to have near perfect accuracy and fine tuning is not needed. On the other hand, really complex tasks are really difficult to fine-tune and clevem data collection might be pretty expensive. Fine-tuning can help with the use cases somewhere in the midd…

What would you say is an example of one of those “middle” tasks it can help with?

An example I just found worked very well with fine-tuning: I wanted to extract any frame that contained a full-screen presentation slide from a various videos I've archived, only when it's full-screen, and also not capture videos, and some other constraints.

Naturally I reached for CLIP+ViT which got me a ~60% success rate out of the box. Then based on that, I created a tiny training script that read `dataset/{slide,no_slide}` and trained a new head based on that. After adding ~100 samples of each, the success rate landed at 95% which was good enough to call it done, and circle back to iterate once I have more data.

I ended up with a 2.2K large "head_weights.safetensors" that increased the accuracy by ~35% which felt really nice.

Re: The case for the return of fine-tuning

#17

Post by funfunfunction »

Creator of inference.net / schematron here.

There is growing emphasis on efficiency as more companies adopt and scale with LLMs in their products.

Developers might be fine paying GPT-5-Super-AGI-Thinking-Max prices to use the very best models in Cursors, but (despite what some may think about Silicon Valley), businesses do care about efficiency.

And if you can fine-tune an 8b-parameter Llama model on GPT-5 data in < 48 hours and save $100k/mo, you're going to take that opportunity.

Re: The case for the return of fine-tuning

#18

Post by alansaber »

meander_water wrote: post #10

A couple of examples I have seen recently which makes me agree with OP: - PaddleOCR, a 0.9B model that reaches SOTA accuracy across text, tables, formulas, charts & handwriting. [0] - A 3B and 8B model which performs HTML to json extraction at GPT-5 level accuracy at 40-80x less cost, and faster inference. [1] I think it makes sense to fine tune when you're optimizing for a specific task. [0] https://huggingface.co/p…

This comes back to the SLM vs LLM debate (sizes in relative terms), where an SLM can be optimised for a specific task, and out-perform an LLM. But it's not worth it (time, effort) for most tasks unless 1. they are very sensitive to precision or 2. it is ultra-high volume.

Re: The case for the return of fine-tuning

#19

Post by coldtea »

matusp wrote: post #6

Fine-tuning is a good technique to have in a toolbox, but in reality, it is feasible only in some use cases. On one hand, many NLP tasks are already easy enough for LLMs to have near perfect accuracy and fine tuning is not needed. On the other hand, really complex tasks are really difficult to fine-tune and clevem data collection might be pretty expensive. Fine-tuning can help with the use cases somewhere in the midd…

>Fine-tuning is a good technique to have in a toolbox, but in reality, it is feasible only in some use cases.

Yes, 100s of housands of them

Post reply on HN