Get the latest tech news

Show HN: Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training


I replicated Ng's RYS method and found that duplicating 3 specific layers in Qwen2.5-32B boosts reasoning by 17% and duplicating layers 12-14 in Devstral-24B improves logical deduction from 0.2...

None

Get the Android app

Or read this on Hacker News

Read more on:

Photo of training

training

Photo of LLM

LLM

Photo of layers

layers

Related news:

News photo

Nvidia says it can shrink LLM memory 20x without changing model weights

News photo

How LinkedIn replaced five feed retrieval systems with one LLM model, at 1.3 billion-user scale

News photo

Show HN: Context Gateway – Compress agent context before it hits the LLM