4 ms·
Check out “Recursive Language Models”, or RLMs. I believe this method works well because it turns a long context problem (hard for LLMs) into a coding and reas
by dbreunig 7mo ago
Check out “Recursive Language Models”, or RLMs.
I believe this method works well because it turns a long context problem (hard for LLMs) into a coding and reasoning problem (much better!). You’re leveraging the last 18 months of coding RL by changing you scaffold.
- koakuma-chan 7mo agoThis seems really weird to me. Isn't that just using LLMs in a specific way? Why come up with a new name "RLM" instead of saying "LLM"? Nothing changes about the model.
- vimda 7mo agoRLMs are a new architecture, but you can mimic an RLM by providing the context through a tool, yes
- anonymousd3vil 7mo agoNew architecture to building agent, but not the model itself. You still have LLMs, but you kinda give this new agentic loop with a REPL environment where the LLM can try to solve the problem more programmatically.
- dbreunig 7mo ago"Think step by step," was just a sentence you appended to your prompt. It ended up kicking off reasoning training which enabled the massive gains in coding, tool use, and more over the last 18 months. So yeah, it's "just using LLMs in a specific way."