What’s the most effective way to add new domain knowledge into an open LLM? A new blog post from my team covers experiments we did at the beginning of the year to start answering this question. It starts, unsurprisingly, with sweeping your learning rate… www.databricks.com/blog/charact...

When you fail to parse your data that’s a jsonl

Mat is not on 🦋—posting on his behalf!

It's time to revisit common assumptions in IR! Embeddings have improved drastically, but mainstream IR evals have stagnated since MSMARCO + BEIR.

We ask: on private or tricky IR tasks, are rerankers better? Surely, reranking many docs is best?

How many documents should you retrieve when using a reranker? The answer might surprise you!

Check out the excellent work from our intern Mathew on this important retrieval question. 👏

I love the smell of providing executives with actionable insights in the morning

"Son, we live in a world that has dashboards, and those dashboard have to be guarded by data engineers with Spark."

Time for some shitposting about dashboards, since that’s in the zeitgeist:

I call many years ago an eager IT manager telling me we should deprecate all of our traditional SSIS reports. Because all that info could be revealed in a dashboard. And dashboards allow “insight discovery” which is 🔥 …

Any data people in New York want to grab bagels and talk about AI? We have a group that meets every other Thursday morning. I’d be happy to add people to the google group.

text2sql is coming, look busy

you either die making good model, or live long enough to see yourself make bad model go fast

It is surprisingly easy to accidentally dress as shaggy from Scooby-Doo