this post was submitted on 01 Nov 2023
1 points (100.0% liked)

Machine Learning

1 readers
1 users here now

Community Rules:

founded 1 year ago
MODERATORS
 

Paper link http://arxiv.org/abs/2310.18338

Description We introduce DaSLaM, which uses a decomposition generator to decompose complex problems into subproblems that require fewer reasoning steps. These subproblems are answered by a solver. We use a relatively small (13B parameters) LM as the decomposition generator, which we train using policy gradient optimization to interact with a solver LM (regarded as black-box) and guide it through subproblems, thereby rendering our method solver-agnostic. Evaluation on multiple different reasoning datasets reveal that with our method, a 175 billion parameter LM (text-davinci-003) can produce competitive or even better performance, compared to its orders-of-magnitude larger successor, GPT-4. Additionally, we show that DaSLaM is not limited by the solver's capabilities as a function of scale; e.g., solver LMs with diverse sizes give significant performance improvement with our solver-agnostic decomposition technique.

top 2 comments
sorted by: hot top controversial new old
[โ€“] Single_Ring4886@alien.top 1 points 1 year ago

This is thing I was talking about some months ago after I first saw GPT4. Glad someone actually worked in this direction as ideas are somehow cheap these days :)

Found 2 relevant code implementations for "Small Language Models Fine-tuned to Coordinate Larger Language Models improve Complex Reasoning".

Ask the author(s) a question about the paper or code.

If you have code to share with the community, please add it here ๐Ÿ˜Š๐Ÿ™

--

To opt out from receiving code links, DM me.