News for hackers. Stuff that matters.
Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu... by nullc on Saturday February 01, 2025 @03:56AM (#42896967)
requires a backwards trained llm no?

I don't think anyone has pretrained a remotely-close-to-SOTA sized backwards model.

Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu... by seanhunter on Saturday February 01, 2025 @05:17AM (#42897340)
Haven’t read the paper yet just the abstract, but it sounds like it uses a backwards trained llm itself to generate prompts and examples but can do the autodiff on any llm.
Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu... by meame2010 on Saturday February 01, 2025 @01:53PM (#42901028)
We use gpt4o as the backward model. But I’m excited to try deepseek r1 as it has explicit reasoning available.

We are continuously adding more benchmarks to the paper with UTAustin.