Michael Carbin
@mcarbin.bsky.social
Associate Professor in EECS at
@MIT | Principal Scientist at @Databricks | Founding Advisor at @mosaicml | Programming Systems | Neural Networks | Approximate Computing
@MIT | Principal Scientist at @Databricks | Founding Advisor at @mosaicml | Programming Systems | Neural Networks | Approximate Computing
Reposted by Michael Carbin
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...
Characterizing Datasets and Building Better Models with Continued Pre-Training
www.databricks.com
November 25, 2024 at 11:29 PM
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...
Reposted by Michael Carbin
Attention🚨 We are looking for motivated students and researchers to be members of the PLDI 2025 Artifact Evaluation Committee. This year, we are accepting self-nominations (the form is here: forms.gle/2TPmixasDmqM...). Deadline: Dec 23rd, 2024.
For more info: pldi25.sigplan.org/track/pldi-2...
For more info: pldi25.sigplan.org/track/pldi-2...
PLDI 2025 Artifact Evaluation Committee Self Nomination
This form allows any member of the community to nominate *yourself* to be part of the Artifact Evaluation Committee for PLDI 2025. While we cannot select all qualified candidates, we will do our best ...
forms.gle
November 22, 2024 at 11:37 PM
Attention🚨 We are looking for motivated students and researchers to be members of the PLDI 2025 Artifact Evaluation Committee. This year, we are accepting self-nominations (the form is here: forms.gle/2TPmixasDmqM...). Deadline: Dec 23rd, 2024.
For more info: pldi25.sigplan.org/track/pldi-2...
For more info: pldi25.sigplan.org/track/pldi-2...
Reposted by Michael Carbin
If you are looking for a quick way to follow everyone you used to follow on the vile site, there’s a solution! “Sky Follower Bridge” - works like a charm. Make sure to give some monetary love to the developer!
lifehacker.com/tech/use-thi...
lifehacker.com/tech/use-thi...
Use This Extension to Find All Your X Followers on Bluesky
If you're looking to finally leave Elon Musk's X, Bluesky is one of your best alternatives—and Sky Follower Bridge is a web extension that will easily help you find all of your X followers there.
lifehacker.com
November 20, 2024 at 3:05 PM
If you are looking for a quick way to follow everyone you used to follow on the vile site, there’s a solution! “Sky Follower Bridge” - works like a charm. Make sure to give some monetary love to the developer!
lifehacker.com/tech/use-thi...
lifehacker.com/tech/use-thi...
Mat: are rerankers supposed to do this?
Team: 👀
<2 months later>
Paper!
This has been incredibly fun work to be a part; my favorite kind of science is finding holes in commonly held assumptions.
Team: 👀
<2 months later>
Paper!
This has been incredibly fun work to be a part; my favorite kind of science is finding holes in commonly held assumptions.
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?
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?
November 21, 2024 at 5:28 PM
Mat: are rerankers supposed to do this?
Team: 👀
<2 months later>
Paper!
This has been incredibly fun work to be a part; my favorite kind of science is finding holes in commonly held assumptions.
Team: 👀
<2 months later>
Paper!
This has been incredibly fun work to be a part; my favorite kind of science is finding holes in commonly held assumptions.