Thanks for sharing your work. Without rerunning the experiment, could we take a look at the results on ARC-AGI 2, in particular which tasks the model was able to solve and which ones it failed?
This approach reminds me of the meta-learning approach referenced by Joscha Bach in his first interview in the Lex Friedman podcast. He mentions that neural networks are an algorithm that automatically looks for an algorithm that implements the problem. Meta-learning is a level above that and finds an algorithm that discovers a learning algorithm for the given domain, which he claims is closer to how our brain works.
Now this strikes me as an insightful and brilliant observation:
> It's as if humans stored a compressed kernel of deduction and logic that we call upon for everything, while LLMs store this kernel fragmented across domain-specific embeddings. They're overfitting to domain-specific reasoning patterns.
Supplying grid representations in base64, ascii and python nested lists improved the result (from your original approach) , showing that vision reasoning is further fragmented across sub-domains. There is not even a central vision reasoning kernel!
All we're waiting on is the one true unified multi-modal representation space.
eerily similar to DSPy's GEPA
https://arxiv.org/abs/2507.19457
Thanks for sharing your work. Without rerunning the experiment, could we take a look at the results on ARC-AGI 2, in particular which tasks the model was able to solve and which ones it failed?
This approach reminds me of the meta-learning approach referenced by Joscha Bach in his first interview in the Lex Friedman podcast. He mentions that neural networks are an algorithm that automatically looks for an algorithm that implements the problem. Meta-learning is a level above that and finds an algorithm that discovers a learning algorithm for the given domain, which he claims is closer to how our brain works.
they are all actually Recursive Emergence https://github.com/Recursive-Emergence/RE/blob/main/chapter_2_mathematical_foundations.md
++ Good Post, Also, start here stock market, AI research, Crash Courses, 100+ Most Asked ML System Design Case Studies and LLM System Design
Stock Market
https://open.substack.com/pub/stockmarketanalysis04/p/important-stock-market-post-04-which?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/stockmarketanalysis04/p/important-stock-market-analysis-which?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/stockmarketanalysis04/p/important-stock-market-post-02-understand?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/stockmarketanalysis04/p/important-stock-market-post-03-this?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
Crash Courses
https://open.substack.com/pub/crashcourses/p/crash-course-02-a-complete-crash?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/crashcourses/p/crash-course-01-a-complete-crash?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
AI/ML Research
https://open.substack.com/pub/airesearch04/p/ai-research-2-kimi-k2-thinking-a?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/airesearch04/p/ai-research-1-the-transformer-revolution?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
LLM System Design
https://open.substack.com/pub/naina0405/p/most-important-llm-system-design-b31?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://naina0405.substack.com/p/launching-llm-system-design-large?r=14q3sp
https://naina0405.substack.com/p/launching-llm-system-design-2-large?r=14q3sp
[https://open.substack.com/pub/naina0405/p/llm-system-design-3-large-language?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/naina0405/p/important-llm-system-design-4-heart?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
System Design
https://open.substack.com/pub/naina0405/p/system-design-tech-case-study-pulse-862?r=14q3sp&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
https://open.substack.com/pub/naina0405/p/system-design-tech-case-study-pulse-b3c?r=14q3sp&utm_campaign=post&utm_medium=web
https://open.substack.com/pub/naina0405/p/system-design-tech-case-study-pulse-135?r=14q3sp&utm_campaign=post&utm_medium=web
https://open.substack.com/pub/naina0405/p/system-design-tech-case-study-pulse-007?r=14q3sp&utm_campaign=post&utm_medium=web
Just like a charm..looks like GP
Now this strikes me as an insightful and brilliant observation:
> It's as if humans stored a compressed kernel of deduction and logic that we call upon for everything, while LLMs store this kernel fragmented across domain-specific embeddings. They're overfitting to domain-specific reasoning patterns.
Supplying grid representations in base64, ascii and python nested lists improved the result (from your original approach) , showing that vision reasoning is further fragmented across sub-domains. There is not even a central vision reasoning kernel!
All we're waiting on is the one true unified multi-modal representation space.
Awesome work!! Thanks for chatting with me this morning. We also found that the LLMs are able to do better optimizations with English instructions vs. with the code: https://github.com/promptdriven/pdd/blob/main/docs/whitepaper_with_benchmarks/whitepaper_w_benchmarks.md
Thank you for sharing this. Brilliantly simple and elegant solution.
👍