Website Source: blog / cadence_blog
Summary
Pending synthesis from local website source.
Original source title: Cadence - Evolving Code with LLMs for NP-Hard Problems
Extracted Preview
Note: Get nerd sniped for the love of god! Also, if you like this kind of work, and your organization does something similar, consider hiring me? I'm aggressively looking for a job. Would love to [chat](mailto:ysrivastava82@gmail.com)
Cadence - Evolving Code with LLMs for NP-Hard Problems
1. [Homepage](https://github.com/yash-srivastava19/cadence)
2. [Docs](https://cadence.readthedocs.io/en/latest/)
Introduction
There are very few paper that have moved me as much as [AlphaEvolve](https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/AlphaEvolve.pdf). The paper improved on the RL pilled frontier of the current AI research landscape, and proved - by discovering novel algorithms, just how much we can still squeeze performance out of LLMs by improving solutions to hard problems and evolving them to discover new solutions. Here's what [@karpathy](https://x.com/karpathy/status/1944435412489171119) has to say about it, and is more or less what AlphaEvolve does:
>Scaling up RL is all the rage right now. I'm fairly certain RL will continue to yield more intermediate gains, but I also don't expect it to be the full story.
>...
>There's significantly more bits of supervision we extract per rollout via a review/reflect stage of human mechanism of improvement along the lines of "what went well? what didn't go so well? what should I try next time?" etc. and the lessons from this stage feel explicit, like a new string to be added to the system prompt for the future, optionally to be distilled into weights (/intuition) later a bit like sleep.
>...
>Example algorithm: given a task, do a few rollouts, stuff them all into one context window (along with the reward in each case), use a meta-prompt to review/reflect on what went well or not to obtain string "lesson", to be added to system prompt (or more generally modify the current lessons database). Many blanks to fill in, many tweaks possible, not obvious.
When I was going through the paper, something struck to me. The paradigms introduced in AlphaEvolve can be used make progress for optimization problems, where there is no exact "algorithmic breakthrough", but rather heuristic or cleverness breakthrough which, in my humble opinion is something that can't be learned through pattern recognition prevalent in LLMs.
Integration Notes
- Source section:
blog - Local source:
/home/yashs/Desktop/Programming/yash_blog/yash-srivastava19.github.io/blog/cadence_blog.md - Raw copy:
raw/website/yash-srivastava19-github-io/blog/cadence_blog.md