Most of the systems conferences have artifact evaluation process that is done upon paper acceptance.
www.acm.org/publications...
I've been part of the artifact evaluation boards, and it verifies if the system exists as written in the paper, and also the performance.
It is still a volunteer based system, and could act as a supplement to the arxiv model of releasing papers. Till date, papers accepted at SOSP/OSDI get much higher citations, than something released on Arxiv.
Yes, to a large extent it is solved by the peer review system.
I'm in systems, take papers from pre-double blind era where the papers were from select instituitions, now the instituitional diversity of the papers getting accepted has widened.
This is not to say peer review doesn't have problems.
There'll always be outstanding works from lower instituitions that get the same fame, but more often than not it is harder to get buy-in for your work if you already don't have a base of viewers.
The main issue is that works from lower ranked universities or newer researchers take a long time to bubble up in this system. If something comes from Berkeley it gets more eyeballs than say something from a non T20 school. Peer review acts as a validator for smaller venues to get recognized.
LEWIS HAMILTON WINS #F1SPRINT, HIS FIRST VICTORY FOR FERRARI 🤩
#F1 #ChineseGP
This is amazing work! I have been facing this problem in my log search project where all the queries are time sensitive. Eventually we had to use a "query planner" llm that converted original requests timestamps to a more machine readable format and then converting it to unix timestamp.
Not an internship, but GSoC applications start on March 24th.
summerofcode.withgoogle.com/get-started
Not fully research role, but Meta has been hiring a lot of Systems/Compiler folks recently.
www.metacareers.com/jobs/9568970...
Yes, it doesn't support logs natively. But you can emit logs in the OpenTelemetry format and have it stored in the UI separately. Traces and Logs are kinda distinct problems though, so you'd need to find another way of storing any querying these logs. Happy to chat more if you want to know how!
Jaeger has a really nice UI module, but the ingesting format has to be OpenTelemetry compatible.
github.com/jaegertracin...
They even played Kendrick's Humble in the stadium after the game!
As someone in grad school, the funding freeze hasn't really been averted. Some of my friends on federal grants have been told their projects are on hold, pending review. I don't understand how he's been stopped on any of this currently.
There's no solid recommendation algorithm yet, so you kinda need to follow the right people! I am down to put a list of US Systems (broadly interpreted) Researchers to have a custom feed.
The dude paused all federal grant payouts, and for 2-3 days PIs didn't know if they'd have money to keep funding their labs. Also, how they'd apply for funding to continue research as well.
I really like OSTEP by Andrea and Remzi!
Anyone recently switched from iPhone to Pixel?
I am debating between iPhone 16 and Pixel 9 Pro because Pixel seems to be much better in terms of software, but don't know about its hardware long term.
Having a huge TV as monitor has a lot of perks when making presentations
thesis pages written: 0
title puns made: inf
When you think you’re so down, you didn’t even know you had a we are so back in you!
In the last 3 months, I read a lot of papers which are the same-ish architecture applied to a dataset, each claiming to have beaten SOTA published at top venues in the last 3 years. There's significant gap between deadlines!
How do reviewers or even authors keep track of such duplicate work?
Feel the same about writing academic papers as well! The more I wrote, the better I got at it!
One peculiar thing I’ve noticed when using LLMs is that the more you try to understand why something happens, the less you actually understand. Instead, approaching them with an open mind and focusing on getting tasks done rather than dissecting how they work seems to be a more effective strategy.
Tbf if the environment was this red, then having Shapiro would have hurt him in a future presidential bid. I seriously think 2028 would be between him or Gallego, I lean more towards Gallego as Rust Belt might be harder to swing than Sun Belt.
Delta has free WiFi onboard for the whole flight duration
Apart from the need to maintain GPUs and high inference latencies for LLMs, classical ML is much better when the data is structured which is true for a lot of low level systems use cases. Performance of deep learning models for tabular data is still not on par with XGBoost, for example.
The current me thanks the past me's decision to have a really effective structure for the latex projects. It took me less than 1 afternoon to port everything into thesis template.
Likewise, looking forward to your defense!!
Very early given I’ve my pre defense in 2 weeks’ time!
(The papers are written, need to write glue sections)
Overleaf went down because it was so surprised that I started writing my thesis that I'd promised to write for the last 6 months.