The case against “scaling is all you need” for LLMs
A commenter argues that large language models needed better data, optimized prompts and other techniques—not just scaling.
TLDR
In a reply, a commenter challenges the “scaling is all you need” argument they say they heard from people working on LLMs. They argue that better data, reinforcement learning with verifiable rewards (RLVR), optimized prompts and a “clever harness” were also needed. They also speculate that substantial unpublished work contributed.
Combined views
1 Source, first seen 3h ago
1 reposts