cowhi

[Brian Bartoldson @bartoldson]

18 months before the Hugging Face hack:

"On January 7 and 8, 2025, several Federally Funded Research and Development Centers (FFRDCs) organized a small, focused, cross-lab workshop to consider how to assess and monitor the possibility that rogue AI could present an existential threat."

Now on arXiv.
https://t.co/t7mc46ndgv

[Charlie London @CharlieLondon02]

We've identified 21 unsolvable problems in LongCoT-mini maths (21 of 507 mini questions), traced to missing information in some subproblems. This error does not affect other questions. Affected question IDs: https://t.co/TSneIMTyTy.

We're working on a fix, ships very soon!

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

LongCoT: Benchmarking Long-Horizon Chain-of-Thought Reasoning

"We introduce LongCoT, a scalable benchmark of 2,500 expert-designed problems spanning chemistry, mathematics, computer science, chess, and logic to isolate and directly measure the long-horizon CoT reasoning capabilities of frontier models."

"At release, the best models achieve <10% accuracy (GPT 5.2: 9.8%; Gemini 3 Pro: 6.1%) on LongCoT, revealing a substantial gap in current capabilities."

[Sumeet Motwani @sumeetrm]

We’re releasing LongCoT, an incredibly hard benchmark to measure long-horizon reasoning capabilities over tens to hundreds of thousands of tokens.

LongCoT consists of 2.5K questions across chemistry, math, chess, logic, and computer science. Frontier models score less than 10%🧵

[Sumeet Motwani @sumeetrm]

Very excited to announce HorizonMath with @erikyw26 and collaborators!

How can we measure AI progress on mathematical discovery? Turns out there’s several classes of problems where discovery is hard but verification is easy. We develop a benchmark with 101 such problems and test GPT 5.4 Pro, Claude 4.6 Opus, and Gemini 3.1 Pro.

Pending expert review, GPT 5.4 Pro finds two potentially novel solutions that beat existing baselines🧵

[Aran Komatsuzaki @arankomatsuzaki]

LongCoT: Benchmarking Long-Horizon Chain-of-Thought Reasoning

- 2,500 expert-designed problems spanning science, chess, and logic that demand up to hundreds of thousands of reasoning tokens
- The best models achieve &lt;10% accuracy

[Elon Musk @elonmusk]

Sorry, too many options. Will redo poll.

[Mikel Landajuela @Landajuela_M]

Paper: https://t.co/AdutsuV7SB
Code: https://t.co/bcJzf4ZyFw
Poster Session #ICML21: https://t.co/Gw3VNdJone

[Mikel Landajuela @Landajuela_M]

@Livermore_Lab, we have developed a #DeepLearning method to discover simple symbolic policies for #ReinforcementLearning. In our work @icmlconf, we show that our models can outperform highly-tuned, pre-trained #NeuralNetworks-based policies, even though they are much more simple!

[cowhi @cowhi_org]

@BinanceUShelp User ID 52399245, stuck in address verification over 3 weeks now. Tried to make a ticket, got no ticket number. What can I do? Please help!

[cowhi @cowhi_org]

@CBSAllAccess how long does it take for the @UCLonCBSSports match replays to be available? Why not immediately??

[PAPIs.io @papisdotio]

Last days of ticket sales, 5 days until the #conference starts! Real-world #machinelearning #stories at https://t.co/tPMUigVquU #latam 2019. Program: https://t.co/m7H1ivD0X3

#ml #ai #data #bigdata #analytics #datascience #dl #technology #ds #tensorflow #deeplearning #networking

[PAPIs.io @papisdotio]

📢 We are happy to announce @TechGrupoZAP as PLATINUM💎 sponsor for #papisio !

https://t.co/Efu6bPSP2X

#ml #ai #artificialintelligence #data #bigdata #analytics #datascience #dl #technology #ds #tensorflow #deeplearning #python #networking #neuralnetworks

[PAPIs.io @papisdotio]

Only a few spots left to support our applied #machinelearning #community! Become a #sponsor and present your company at the https://t.co/tPMUihd1ms LATAM 2019 #conference on June 24-26 in São Paulo. Contact us through https://t.co/qbFNh6IuHr

[PAPIs.io @papisdotio]

Very excited to announce that Guilherme Silveira @guilhermecaelum, Head of Education and Innovation @AluraOnline will be talking at #papisio LATAM about bias and bugs on recommendation systems.

Check the conference program at: https://t.co/m7H1ivD0X3

[PAPIs.io @papisdotio]

Bianca Zadrozny @biancazad , Manager of Natural Resources Analytics Research at @ibmbrasil , will introduce Machine Learning for subsurface characterization as featured speaker at #papisio

Check the conference program at https://t.co/4SWxjRDfOe and follow us for more news!

[LatinX in AI (LXAI) @_LXAI]

Did you catch last year's @papisdotio Conference in Brazil? Join them again on June 24th-26th, 2019!
#AI #LATAM #LatinXinAI
https://t.co/ynHwMlkjjB

[andryw @andryw]

My Talk at PAPIs LATAM 2019 - Using Machine Learning to recommend jobs in User Cold Start scenarios
#papisio #machinelearning #ml #ai #datasciene #conference @papisdotio

[PAPIs.io @papisdotio]

We are excited to announce Piero Molino @w4nderlus7 as featured speaker at https://t.co/tPMUigVquU LATAM, where he will present Ludwig, a code-free #DeepLearning toolbox

Check the conference program at https://t.co/4SWxjRDfOe and @w4nderlus7 talk at https://t.co/g0vDAxruMQ

[PAPIs.io @papisdotio]

Delighted to announce Moussa Taifi as #papisio featured speaker. Moussa is a Senior Data Science Platform Engineer at @Xandr76, and will present a scheduling mechanism for #ML batch jobs

Conference program: https://t.co/4SWxjRDfOe and Moussa's talk at https://t.co/tBTMjsrwmb