768
La
learning-ai-lua-code-model / files
learning-ai-lua-code-model / notes / 008-future-rlm-harness.md
1.3 KB
1# Future experiment: RLM-style harness 2 3This is a reminder for a much later experiment inspired by the Recursive Language Models paper. 4 5## Idea 6 7Build a small educational harness around one of our Lua language models or a stronger external/local model where: 8 9```text 10large input corpus lives outside the model context 11model/sampler receives metadata and selected snippets 12code handles chunking, looping, storage, and aggregation 13semantic work happens through model calls 14``` 15 16The point is to explore the systems idea: 17 18```text 19do not load everything into the model context; 20keep a handle to the data and process it in chunks 21``` 22 23## Possible toy version 24 25Use a Lua corpus or documentation bundle and ask questions like: 26 27- find sections about tables/metatables 28- summarize all examples of functions 29- collect idioms from many small Lua files 30- compare generated summaries from chunked vs direct prompts 31 32## Learning goals 33 34- understand context windows as limited working memory 35- practice chunking and map/reduce-style processing 36- separate exact code operations from fuzzy model calls 37- see where recursion/sub-calls help or become expensive 38- connect classic software memory-management intuition to modern LLM systems 39 40This is intentionally a future milestone. We should first keep improving the basic training, sampling, evaluation, and distillation experiments.