serendip-ml

5 ms·

1. 1.▲
   by serendip-ml 1mo ago

   Interesting question is if the CoE Audit evaluates evidence chains recursively (each cited paper's own evidence chain), or stops at citation existence + method-code alignment for the current paper. For example, citing a paper whose met
2. 2.▲
   Show HN: Fine-tuned 3B outperforms Claude Haiku on constrained generation(serendip-ml.github.io)

   1 points by serendip-ml 6mo ago|0 comments
3. 3.▲
   by serendip-ml 7mo ago

   Author here. Curious what others think about the verb-based approach vs pure schema validation. Happy to discuss trade-offs.
4. 4.▲
   Show HN: SAIA – Typed Contracts for LLM Outputs(github.com)

   2 points by serendip-ml 7mo ago|1 comments
5. 5.▲
   by serendip-ml 7mo ago

   Hi HN, Karpathy's recent post [1] described Claws as "a new layer on top of LLM agents, taking orchestration, scheduling, context, tool calls to a next level." That's the right framing - but orchestration alone isn'
6. 6.▲
   Show HN: SAIA – SCUMM for AI Agents(github.com)

   3 points by serendip-ml 7mo ago|1 comments
7. 7.▲
   by serendip-ml 7mo ago

   The compression analogy is interesting. Another way of looking at it could be fine-tuning as "knowing what to leave out" - a 3B model for example tuned for a narrow task doesn't need the capacity that makes 70B good at many t