GenAI for Systems. Don’t fight it, just embrace it! #mlforsystems

MimeeXu's tweet image. GenAI for Systems. Don’t fight it, just embrace it!

#mlforsystems

NeurIPS 2025 WS on ML for Systemsに論文が2本採択されました! #MLforSystems #NeurIPS2025 (1) 去年卒論の西下君、D1佐藤君による学習型カウントミンスケッチ (2) M1の山下君、阪大 @Daichi_Amagata 先生によるANN削除方式の評価


Tomorrow at NeurIPS: MACHINE LEARNING FOR SYSTEMS @JeffDean opens at 8:20AM, so come early (west 201) The most anticipated workshop, for us anyway! @tim_kraska @natashajaques @ahelkky #mlforsystems @DZhang50 @haoran_qiu98 @divyamahajn @OpenAI @MSFTResearch @GoogleDeepMind

MimeeXu's tweet image. Tomorrow at NeurIPS:

MACHINE LEARNING FOR SYSTEMS

@JeffDean opens at 8:20AM, so come early (west 201)

The most anticipated workshop, for us anyway!

@tim_kraska @natashajaques @ahelkky 

#mlforsystems
@DZhang50 @haoran_qiu98 @divyamahajn 
@OpenAI @MSFTResearch @GoogleDeepMind

We are paneling! Our own Mangpo is moderating. Submit your questions. #MLforSystems @natashajaques @JeffDean Lidong Zhou and @tim_kraska

MimeeXu's tweet image. We are paneling! Our own Mangpo is moderating.

Submit your questions. #MLforSystems

@natashajaques @JeffDean Lidong Zhou and @tim_kraska

Excited to announce our first website refresh in the last 6 years! monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago. We have matured! #MLforSystems #neurips2024

MimeeXu's tweet image. Excited to announce our first website refresh in the last 6 years!

monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago.

We have matured! #MLforSystems #neurips2024

ML models need compute but Moore’s Law is slowing down. Are flops enough? Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. per watts, perf, per TCO. #GPUs #TPUs #MLforSystems #NeurIPS2024

MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024
MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024

We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times. And it won a gold. #CodeGen #MLforSystems #openaio1 @ahelkky

MimeeXu's tweet image. We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times.

And it won a gold.

#CodeGen #MLforSystems #openaio1
@ahelkky

Data may have run out but systems can be faster. Or can they? #mlforsystems @JeffDean West Room 201. (You may have to stand because the room is getting full)

MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)
MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)

ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

haoran_qiu98's tweet image. ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

#MLforSystems Works in progress! Our spotlights: Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization Accepted papers: mlforsystems.org/accepted_paper…

MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…

Two papers have been accepted to the NeurIPS 2025 Workshop on ML for Systems! #MLforSystems #NeurIPS2025 (1) Learned Count-Min Sketch by Kyosuke and Atsuki (2) Evaluation of ANN Deletion by Tomohiro and Prof. Amagata @Daichi_Amagata (UOsaka)


In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. #riscv #hardware #mlforsystems Open sourced! Retrained checkpoint released! @Azaliamirh @annadgoldie

MimeeXu's tweet image. In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. 

#riscv #hardware #mlforsystems 

Open sourced! Retrained checkpoint released!

@Azaliamirh @annadgoldie

Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open. Submission deadline: Aug 22 AoE Help spread the word! P.S. Agents and LLM systems are systems, too! mlforsystems.org/call_for_paper… #MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

MimeeXu's tweet image. Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open.

Submission deadline: Aug 22 AoE

Help spread the word!
P.S. Agents and LLM systems are systems, too!

mlforsystems.org/call_for_paper…
#MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

PSA: our whole event stream is available through NeurIPS’s official site: Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction. neurips.cc/virtual/2023/w… #MLforSystems

MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems

If you missed the NeurIPS ML for Systems Workshop, the presentations are recorded and available here: slideslive.com/neurips-2023/w… #mlforsystems #neurips



M1日髙君の論文がNeurIPS WSのML for Systemsに採択されました!学習型マルチインデックスをデータ更新に頑健にしました。 #MLforSystems #NeurIPS2024 mlforsystems.org


After releasing the original ComPile dataset, we are excited to present an update on the dataset, and its public version later today at the ML 4 Systems workshop at NeurIPS. Please stop by our poster, or reach out if you want to chat! #MLforSystems #NeurIPS2023 #MLSys

ludgerpaehler's tweet image. After releasing the original ComPile dataset, we are excited to present an update on the dataset, and its public version later today at the ML 4 Systems workshop at NeurIPS. Please stop by our poster, or reach out if you want to chat!

#MLforSystems #NeurIPS2023 #MLSys

Huge thanks to my advisor @kkaffes, and @DAP__Lab for the support on our work! 🙏🎉 #OperatingSystems #MachineLearning #MLforSystems #Linux #PhD


Fuma's paper has been accepted to NeurIPS WS on ML for Systems! We proposed a robust data update method for learned multi-indexes. #MLforSystems #NeurIPS2024 mlforsystems.org


Huge thanks to my advisor @kkaffes, and @DAP__Lab for the support on our work! 🙏🎉 #OperatingSystems #MachineLearning #MLforSystems #Linux #PhD


Two papers have been accepted to the NeurIPS 2025 Workshop on ML for Systems! #MLforSystems #NeurIPS2025 (1) Learned Count-Min Sketch by Kyosuke and Atsuki (2) Evaluation of ANN Deletion by Tomohiro and Prof. Amagata @Daichi_Amagata (UOsaka)


NeurIPS 2025 WS on ML for Systemsに論文が2本採択されました! #MLforSystems #NeurIPS2025 (1) 去年卒論の西下君、D1佐藤君による学習型カウントミンスケッチ (2) M1の山下君、阪大 @Daichi_Amagata 先生によるANN削除方式の評価


‼️DEADLINE EXTENDED: ML for Systems @NeurIPSConf now due on 8/29. 🆕So many “agents” talk on X, we wonder how many are rigorous (and scientifically sound!). 📝We are ready to review and give valuable feedback on LLM works, even for first time participants. #mlforsystems #cfp

Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open. Submission deadline: Aug 22 AoE Help spread the word! P.S. Agents and LLM systems are systems, too! mlforsystems.org/call_for_paper… #MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

MimeeXu's tweet image. Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open.

Submission deadline: Aug 22 AoE

Help spread the word!
P.S. Agents and LLM systems are systems, too!

mlforsystems.org/call_for_paper…
#MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025


We are extending the deadline to **Aug 29** I repeat, we are extending the deadline for ML for Systems at NeurIPS to Aug 29 AoE! Submit your 4-page abstract on using ML for system design, including but not limited to coding agents. #mlforsystems #cfp #neurips2025


The 9th edition of the ML for Systems workshop is back at NeurIPS 2025! This year, we’re especially excited to spotlight agentic workflows and generative AI systems. Submit your work before Aug 22 AoE!👇 mlforsystems.org/call_for_paper… #MLforSystems #NeurIPS2025 #LLMs #Agents

Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open. Submission deadline: Aug 22 AoE Help spread the word! P.S. Agents and LLM systems are systems, too! mlforsystems.org/call_for_paper… #MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

MimeeXu's tweet image. Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open.

Submission deadline: Aug 22 AoE

Help spread the word!
P.S. Agents and LLM systems are systems, too!

mlforsystems.org/call_for_paper…
#MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025


Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open. Submission deadline: Aug 22 AoE Help spread the word! P.S. Agents and LLM systems are systems, too! mlforsystems.org/call_for_paper… #MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

MimeeXu's tweet image. Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open.

Submission deadline: Aug 22 AoE

Help spread the word!
P.S. Agents and LLM systems are systems, too!

mlforsystems.org/call_for_paper…
#MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

PSA: our whole event stream is available through NeurIPS’s official site: Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction. neurips.cc/virtual/2023/w… #MLforSystems

MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems

If you missed the NeurIPS ML for Systems Workshop, the presentations are recorded and available here: slideslive.com/neurips-2023/w… #mlforsystems #neurips



The paper I spoke about at the NeurIPS ML for Systems #mlforsystems workshop is now on arxiv arxiv.org/abs/2412.15573! We use multi-agent RL to solve a classic combinatorial optimization problem (the Sequential Assignment Problem) by combining a polynomial time assignment…

My first paper will be in AAAI this yr (@natashajaques) - we figured out how to get 1000s of satellites to play nicely with distributed optimization + RL, smashing baselines. People severely underestimate the maturity level of RL in 2024, and the impacts it will/is already having

joshgholder's tweet image. My first paper will be in AAAI this yr (@natashajaques) - we figured out how to get 1000s of satellites to play nicely with distributed optimization + RL, smashing baselines. People severely underestimate the maturity level of RL in 2024, and the impacts it will/is already having


#MLforSystems Works in progress! Our spotlights: Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization Accepted papers: mlforsystems.org/accepted_paper…

MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…

We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times. And it won a gold. #CodeGen #MLforSystems #openaio1 @ahelkky

MimeeXu's tweet image. We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times.

And it won a gold.

#CodeGen #MLforSystems #openaio1
@ahelkky

We are paneling! Our own Mangpo is moderating. Submit your questions. #MLforSystems @natashajaques @JeffDean Lidong Zhou and @tim_kraska

MimeeXu's tweet image. We are paneling! Our own Mangpo is moderating.

Submit your questions. #MLforSystems

@natashajaques @JeffDean Lidong Zhou and @tim_kraska

GenAI for Systems. Don’t fight it, just embrace it! #mlforsystems

MimeeXu's tweet image. GenAI for Systems. Don’t fight it, just embrace it!

#mlforsystems

ML models need compute but Moore’s Law is slowing down. Are flops enough? Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. per watts, perf, per TCO. #GPUs #TPUs #MLforSystems #NeurIPS2024

MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024
MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024

In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. #riscv #hardware #mlforsystems Open sourced! Retrained checkpoint released! @Azaliamirh @annadgoldie

MimeeXu's tweet image. In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. 

#riscv #hardware #mlforsystems 

Open sourced! Retrained checkpoint released!

@Azaliamirh @annadgoldie

Data may have run out but systems can be faster. Or can they? #mlforsystems @JeffDean West Room 201. (You may have to stand because the room is getting full)

MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)
MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)

Fuma will present our work on the updatable learned multi-dimensional index at the ML for Systems workshop on Dec 15. Feel free to attend! #MLforSystems #NeurIPS2024 arxiv.org/abs/2411.09205


Tomorrow at NeurIPS: MACHINE LEARNING FOR SYSTEMS @JeffDean opens at 8:20AM, so come early (west 201) The most anticipated workshop, for us anyway! @tim_kraska @natashajaques @ahelkky #mlforsystems @DZhang50 @haoran_qiu98 @divyamahajn @OpenAI @MSFTResearch @GoogleDeepMind

MimeeXu's tweet image. Tomorrow at NeurIPS:

MACHINE LEARNING FOR SYSTEMS

@JeffDean opens at 8:20AM, so come early (west 201)

The most anticipated workshop, for us anyway!

@tim_kraska @natashajaques @ahelkky 

#mlforsystems
@DZhang50 @haoran_qiu98 @divyamahajn 
@OpenAI @MSFTResearch @GoogleDeepMind

Excited to announce our first website refresh in the last 6 years! monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago. We have matured! #MLforSystems #neurips2024

MimeeXu's tweet image. Excited to announce our first website refresh in the last 6 years!

monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago.

We have matured! #MLforSystems #neurips2024

ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

haoran_qiu98's tweet image. ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

Tomorrow at NeurIPS: MACHINE LEARNING FOR SYSTEMS @JeffDean opens at 8:20AM, so come early (west 201) The most anticipated workshop, for us anyway! @tim_kraska @natashajaques @ahelkky #mlforsystems @DZhang50 @haoran_qiu98 @divyamahajn @OpenAI @MSFTResearch @GoogleDeepMind

MimeeXu's tweet image. Tomorrow at NeurIPS:

MACHINE LEARNING FOR SYSTEMS

@JeffDean opens at 8:20AM, so come early (west 201)

The most anticipated workshop, for us anyway!

@tim_kraska @natashajaques @ahelkky 

#mlforsystems
@DZhang50 @haoran_qiu98 @divyamahajn 
@OpenAI @MSFTResearch @GoogleDeepMind

GenAI for Systems. Don’t fight it, just embrace it! #mlforsystems

MimeeXu's tweet image. GenAI for Systems. Don’t fight it, just embrace it!

#mlforsystems

Kernel fusion straight to Assembly 🚀🚀🚀 Folks now write fake compilers in Python to hack around constraints -> sad & hard to maintain. Enabling hacking on compilers is truly where the magic is! Chris introduces Modular’s MAX: productive and fast AI Engine. #MLforSystems

MimeeXu's tweet image. Kernel fusion straight to Assembly 🚀🚀🚀

Folks now write fake compilers in Python to hack around constraints ->  sad & hard to maintain.

Enabling hacking on compilers is truly where the magic is!

Chris introduces Modular’s MAX: productive and fast AI Engine.

#MLforSystems

Come to our ML for Systems poster session at 11:35 if y’all have more questions for Chris! #NeurIPS2023 #MLforSystems #Modular @clattner_llvm @DZhang50

MimeeXu's tweet image. Come to our ML for Systems poster session at 11:35 if y’all have more questions for Chris!

#NeurIPS2023 #MLforSystems #Modular 
@clattner_llvm @DZhang50

After releasing the original ComPile dataset, we are excited to present an update on the dataset, and its public version later today at the ML 4 Systems workshop at NeurIPS. Please stop by our poster, or reach out if you want to chat! #MLforSystems #NeurIPS2023 #MLSys

ludgerpaehler's tweet image. After releasing the original ComPile dataset, we are excited to present an update on the dataset, and its public version later today at the ML 4 Systems workshop at NeurIPS. Please stop by our poster, or reach out if you want to chat!

#MLforSystems #NeurIPS2023 #MLSys

🏆🏆 In the end, we all learn so much from people who actually solve systems problems with ML. Many winners didn’t know much about GNNs before! The competition chairs want to thank the community for helping each other and @kaggle for sponsoring part of the prize! #MLforSystems

MimeeXu's tweet image. 🏆🏆 
In the end, we all learn so much from people who actually solve systems problems with ML. Many winners didn’t know much about GNNs before!

The competition chairs want to thank the community for helping each other and @kaggle for sponsoring part of the prize!

#MLforSystems
MimeeXu's tweet image. 🏆🏆 
In the end, we all learn so much from people who actually solve systems problems with ML. Many winners didn’t know much about GNNs before!

The competition chairs want to thank the community for helping each other and @kaggle for sponsoring part of the prize!

#MLforSystems
MimeeXu's tweet image. 🏆🏆 
In the end, we all learn so much from people who actually solve systems problems with ML. Many winners didn’t know much about GNNs before!

The competition chairs want to thank the community for helping each other and @kaggle for sponsoring part of the prize!

#MLforSystems

The rest of NeurIPS: how we can do ML on these soft squishy things like UX (…or “aligning human preference”) if we don’t know how to quantify them? Chrome OS team: ugh.. regression? User experience on Chrome books can be predicted from specs! #MLforSystems #UX #Alignment

MimeeXu's tweet image. The rest of NeurIPS: how we can do ML on these soft squishy things like UX (…or “aligning human preference”) if we don’t know how to quantify them?

Chrome OS team: ugh.. regression?

User experience on Chrome books can be predicted from specs!

#MLforSystems #UX #Alignment
MimeeXu's tweet image. The rest of NeurIPS: how we can do ML on these soft squishy things like UX (…or “aligning human preference”) if we don’t know how to quantify them?

Chrome OS team: ugh.. regression?

User experience on Chrome books can be predicted from specs!

#MLforSystems #UX #Alignment
MimeeXu's tweet image. The rest of NeurIPS: how we can do ML on these soft squishy things like UX (…or “aligning human preference”) if we don’t know how to quantify them?

Chrome OS team: ugh.. regression?

User experience on Chrome books can be predicted from specs!

#MLforSystems #UX #Alignment

We are paneling! Our own Mangpo is moderating. Submit your questions. #MLforSystems @natashajaques @JeffDean Lidong Zhou and @tim_kraska

MimeeXu's tweet image. We are paneling! Our own Mangpo is moderating.

Submit your questions. #MLforSystems

@natashajaques @JeffDean Lidong Zhou and @tim_kraska

ML models need compute but Moore’s Law is slowing down. Are flops enough? Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. per watts, perf, per TCO. #GPUs #TPUs #MLforSystems #NeurIPS2024

MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024
MimeeXu's tweet image. ML models need compute but Moore’s Law is slowing down. Are flops enough?

Come hear Richard’s take on chip and system games and maximizing efficiency across everything: network storage memory and compute. 

per watts, perf, per TCO. 

#GPUs #TPUs #MLforSystems #NeurIPS2024

ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

haoran_qiu98's tweet image. ML for Systems workshop (at #neurips2024) schedule is out! Packed with talks on the latest advancements in machine learning for systems! Please check out! #NeurIPS2024 #MLforSystems

In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. #riscv #hardware #mlforsystems Open sourced! Retrained checkpoint released! @Azaliamirh @annadgoldie

MimeeXu's tweet image. In AlphaChip, pre-training on blocks would give our model intuition, which is what you want to do if you want to do reinforcement learning. 

#riscv #hardware #mlforsystems 

Open sourced! Retrained checkpoint released!

@Azaliamirh @annadgoldie

We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times. And it won a gold. #CodeGen #MLforSystems #openaio1 @ahelkky

MimeeXu's tweet image. We relaxed one of the constraints for OpenAI o-1 to compete in IOI: it could submit 10,000 times.

And it won a gold.

#CodeGen #MLforSystems #openaio1
@ahelkky

Excited to announce our first website refresh in the last 6 years! monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago. We have matured! #MLforSystems #neurips2024

MimeeXu's tweet image. Excited to announce our first website refresh in the last 6 years!

monospace title with purple hues, featuring a beautiful loss landscape @jonathanrraiman whipped up ( i think ) 6 years ago.

We have matured! #MLforSystems #neurips2024

Data may have run out but systems can be faster. Or can they? #mlforsystems @JeffDean West Room 201. (You may have to stand because the room is getting full)

MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)
MimeeXu's tweet image. Data may have run out but systems can be faster. Or can they?

#mlforsystems 

@JeffDean West Room 201. (You may have to stand because the room is getting full)

Dropping tokens improves generative AI? “Almost poetic.” “Empirically, not all tokens are needed.” Come catch Prof. Atlas Wang from @UTAustin on Heavy Hitters policy for sparse attention, with ✨ surprising and unintended ✨ benefits. Room 211-213. #NeurIPS2023 #MLforSystems

MimeeXu's tweet image. Dropping tokens improves generative AI?

“Almost poetic.”

“Empirically, not all tokens are needed.”

Come catch Prof. Atlas Wang from @UTAustin on Heavy Hitters policy for sparse attention, with ✨ surprising and unintended ✨ benefits. 

Room 211-213.
#NeurIPS2023 #MLforSystems
MimeeXu's tweet image. Dropping tokens improves generative AI?

“Almost poetic.”

“Empirically, not all tokens are needed.”

Come catch Prof. Atlas Wang from @UTAustin on Heavy Hitters policy for sparse attention, with ✨ surprising and unintended ✨ benefits. 

Room 211-213.
#NeurIPS2023 #MLforSystems

I am also co-organizing the ML for Systems workshop: mlforsystems.org. We have invited talks from Bill Dally, Chris Lattner, Atlas Wang, and much more. Please join us on Saturday 12/16 and find the schedule at: mlforsystems.org/schedule.html



#MLforSystems Works in progress! Our spotlights: Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization Accepted papers: mlforsystems.org/accepted_paper…

MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…
MimeeXu's tweet image. #MLforSystems

Works in progress! Our spotlights:

Ask LLMs to transform the code, LLMs for optimizing queries, and using agents to do ninja programming optimization 

Accepted papers: mlforsystems.org/accepted_paper…

PSA: our whole event stream is available through NeurIPS’s official site: Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction. neurips.cc/virtual/2023/w… #MLforSystems

MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems
MimeeXu's tweet image. PSA: our whole event stream is available through NeurIPS’s official site:

Bill Dally @nvidia, Chris Lattner @clattner_llvm, Atlas Wang, + Mangpo running the first big @kaggle-sponsored competition on XLA graph to performance prediction.

neurips.cc/virtual/2023/w… #MLforSystems

If you missed the NeurIPS ML for Systems Workshop, the presentations are recorded and available here: slideslive.com/neurips-2023/w… #mlforsystems #neurips



Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open. Submission deadline: Aug 22 AoE Help spread the word! P.S. Agents and LLM systems are systems, too! mlforsystems.org/call_for_paper… #MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

MimeeXu's tweet image. Exciting news! 📣 The call for papers for ML for Systems @NeurIPSConf is now open.

Submission deadline: Aug 22 AoE

Help spread the word!
P.S. Agents and LLM systems are systems, too!

mlforsystems.org/call_for_paper…
#MLforSystems #MachineLearning #LLMs #Agents #CodeModels #neurips2025

Loading...

Something went wrong.


Something went wrong.


United States Trends