Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Poseidon needs voice data and reliable ground truth in low-resource languages to benchmark against. To ensure LLM transcript accuracy, we worked with linguists to audit Bengali outputs. For a language spoken by 280M people, the gaps we found point to a deeper issue: data ↓

Poseidon

25,953 subscribers

22,705 görüntüleme • 4 ay önce •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Every forecast, every warning, and every dataset we share helps countries turn data to action. By closing data gaps in observations, we can ensure that communities are informed, warned early and better protected wherever they may be. Our power, our planet. #EarthDay

Every forecast, every warning, and every dataset we share helps countries turn data to action. By closing data gaps in observations, we can ensure that communities are informed, warned early and better protected wherever they may be. Our power, our planet. #EarthDay

World Meteorological Organization

36,556 görüntüleme • 2 ay önce

The Power of Decentralized Data 💪 Data is the new oil, 🛢️ but it's also a valuable asset that needs to be protected. 🛡️ Traditional centralized data storage and processing methods are vulnerable to security breaches and data leaks. 🚨 Inflectiv's decentralized approach offers a more secure and efficient way to manage data. 🔐 By distributing data across the network, we reduce the risk of data loss and ensure that it remains under the control of its rightful owners. 🤝 Experience the power of decentralized data with Inflectiv. 🚀 #Inflectiv #DecentralizedData #DataSecurity #Blockchain

The Power of Decentralized Data 💪 Data is the new oil, 🛢️ but it's also a valuable asset that needs to be protected. 🛡️ Traditional centralized data storage and processing methods are vulnerable to security breaches and data leaks. 🚨 Inflectiv's decentralized approach offers a more secure and efficient way to manage data. 🔐 By distributing data across the network, we reduce the risk of data loss and ensure that it remains under the control of its rightful owners. 🤝 Experience the power of decentralized data with Inflectiv. 🚀 #Inflectiv #DecentralizedData #DataSecurity #Blockchain

Inflectiv AI ⧉

32,581 görüntüleme • 1 yıl önce

AI was trained on the open internet, but the data that matters most lives in the real world. Introducing early access to Numo, an app built to collect the next generation of AI training data. Starting with voice data collection in Bengali, Hindi, Tamil, and Telugu. Details ↴

AI was trained on the open internet, but the data that matters most lives in the real world. Introducing early access to Numo, an app built to collect the next generation of AI training data. Starting with voice data collection in Bengali, Hindi, Tamil, and Telugu. Details ↴

Poseidon

1,240,151 görüntüleme • 2 ay önce

Picking & choosing “The Data” Welcher & Burgess found JOKs location data reliable when they needed to adjust the time of the triggers because of “clock variance” So why didn’t they trust the location data when trying to figure “where on the road” JOK was when this “collision” happened? Is the location data reliable or not? I thought the data said he was at the flagpole the entire night? Wouldn’t that be where the collision happened since he was found by the flagpole? Did you not want to have to come up with another fake story you had to try and sell?

Picking & choosing “The Data” Welcher & Burgess found JOKs location data reliable when they needed to adjust the time of the triggers because of “clock variance” So why didn’t they trust the location data when trying to figure “where on the road” JOK was when this “collision” happened? Is the location data reliable or not? I thought the data said he was at the flagpole the entire night? Wouldn’t that be where the collision happened since he was found by the flagpole? Did you not want to have to come up with another fake story you had to try and sell?

Joe Flipperhead

20,922 görüntüleme • 1 yıl önce

LongWriter Unleashing 10,000+ Word Generation from Long Context LLMs discuss: Current long context large language models (LLMs) can process inputs up to 100,000 tokens, yet struggle to generate outputs exceeding even a modest length of 2,000 words. Through controlled experiments, we find that the model's effective generation length is inherently bounded by the sample it has seen during supervised fine-tuning (SFT). In other words, their output limitation is due to the scarcity of long-output examples in existing SFT datasets. To address this, we introduce AgentWrite, an agent-based pipeline that decomposes ultra-long generation tasks into subtasks, enabling off-the-shelf LLMs to generate coherent outputs exceeding 20,000 words. Leveraging AgentWrite, we construct LongWriter-6k, a dataset containing 6,000 SFT data with output lengths ranging from 2k to 32k words. By incorporating this dataset into model training, we successfully scale the output length of existing models to over 10,000 words while maintaining output quality. We also develop LongBench-Write, a comprehensive benchmark for evaluating ultra-long generation capabilities. Our 9B parameter model, further improved through DPO, achieves state-of-the-art performance on this benchmark, surpassing even much larger proprietary models. In general, our work demonstrates that existing long context LLM already possesses the potential for a larger output window--all you need is data with extended output during model alignment to unlock this capability.

LongWriter Unleashing 10,000+ Word Generation from Long Context LLMs discuss: Current long context large language models (LLMs) can process inputs up to 100,000 tokens, yet struggle to generate outputs exceeding even a modest length of 2,000 words. Through controlled experiments, we find that the model's effective generation length is inherently bounded by the sample it has seen during supervised fine-tuning (SFT). In other words, their output limitation is due to the scarcity of long-output examples in existing SFT datasets. To address this, we introduce AgentWrite, an agent-based pipeline that decomposes ultra-long generation tasks into subtasks, enabling off-the-shelf LLMs to generate coherent outputs exceeding 20,000 words. Leveraging AgentWrite, we construct LongWriter-6k, a dataset containing 6,000 SFT data with output lengths ranging from 2k to 32k words. By incorporating this dataset into model training, we successfully scale the output length of existing models to over 10,000 words while maintaining output quality. We also develop LongBench-Write, a comprehensive benchmark for evaluating ultra-long generation capabilities. Our 9B parameter model, further improved through DPO, achieves state-of-the-art performance on this benchmark, surpassing even much larger proprietary models. In general, our work demonstrates that existing long context LLM already possesses the potential for a larger output window--all you need is data with extended output during model alignment to unlock this capability.

AK

50,969 görüntüleme • 1 yıl önce

We need to build more data centers. AI needs it.

We need to build more data centers. AI needs it.

Ole

66,388 görüntüleme • 4 ay önce

OVER x Silencio We’re partnering with Silencio | Voice Data for AI to add a new layer of data to the OVER mapping experience. Starting today, users can optionally capture real-world sound while creating their 3D maps inside the OVER app, and we seamlessly share the noise data with Silencio. This will soon allow mappers who capture both visuals and sound to also unlock additional rewards for their contribution. 🌎 We map physical locations around the world. 🌎 Silencio maps the world through sound. Together combining visual + audio signals , we’re enriching the foundation of Physical AI with deeper, multi-layered real-world data.

OVER x Silencio We’re partnering with Silencio | Voice Data for AI to add a new layer of data to the OVER mapping experience. Starting today, users can optionally capture real-world sound while creating their 3D maps inside the OVER app, and we seamlessly share the noise data with Silencio. This will soon allow mappers who capture both visuals and sound to also unlock additional rewards for their contribution. 🌎 We map physical locations around the world. 🌎 Silencio maps the world through sound. Together combining visual + audio signals , we’re enriching the foundation of Physical AI with deeper, multi-layered real-world data.

Over the Reality 🌐

144,078 görüntüleme • 7 ay önce

Drive Success with ThinkerCar's Data-Driven Marketing Solutions 🚀 As a leading provider in the auto industry, we utilize extensive automotive data to provide strategic insights. Our solutions cover data analysis, consulting, and comprehensive marketing support. Quality is our hallmark—partner with us for unparalleled data-driven marketing strategies. Leverage 20+ years of automotive data expertise to make informed decisions. With a track record of serving 200+ B2B clients, we offer customized reports and insights. Gain the competitive edge - reach out to us for top-tier data solutions. 📨 service@thinkercar.com 🌐 🛎️ Telegram: #automotiveindustryinsights #cardata #consulting #marketingstrategy #dataservicessolutions #businessgrowth

Drive Success with ThinkerCar's Data-Driven Marketing Solutions 🚀 As a leading provider in the auto industry, we utilize extensive automotive data to provide strategic insights. Our solutions cover data analysis, consulting, and comprehensive marketing support. Quality is our hallmark—partner with us for unparalleled data-driven marketing strategies. Leverage 20+ years of automotive data expertise to make informed decisions. With a track record of serving 200+ B2B clients, we offer customized reports and insights. Gain the competitive edge - reach out to us for top-tier data solutions. 📨 [email protected] 🌐 🛎️ Telegram: #automotiveindustryinsights #cardata #consulting #marketingstrategy #dataservicessolutions #businessgrowth

ThinkerCar

104,778 görüntüleme • 1 yıl önce

Introducing Databricks LakeFlow, a new solution that simplifies all aspects of data engineering through a simple, unified & intelligent experience. Easily build production-grade data pipelines to scale to the ever-growing demand for reliable #data & #AI.

Introducing Databricks LakeFlow, a new solution that simplifies all aspects of data engineering through a simple, unified & intelligent experience. Easily build production-grade data pipelines to scale to the ever-growing demand for reliable #data & #AI.

Databricks

13,773 görüntüleme • 2 yıl önce

We need a data bill of rights so companies like Palantir don’t give our data to ICE to harass and detain immigrants and American citizens.

We need a data bill of rights so companies like Palantir don’t give our data to ICE to harass and detain immigrants and American citizens.

Ro Khanna

177,786 görüntüleme • 4 ay önce

The newest version of our Almanac preprint is out, and just in time for our demo at the Stanford AIMI Symposium 2023! Almanac is a retrieval-augmented LLM that provides up-to-date and verifiable answers to medical queries. Link: We benchmark our approach on a novel dataset of clinical scenarios (n = 130) evaluated by a panel of 5 board-certified & resident physicians, and demonstrate significant increases in factuality (mean of 18% at p-value < 0.05) across all specialties. More interestingly, because the retrieved data acts as a single source of truth, we find retrieval-based LLMs to be more robust to prompt injection and manipulation! Future work will involve expanding the scope of our dataset to more specialties and multimodal settings. #Medtwitter #MedEd

The newest version of our Almanac preprint is out, and just in time for our demo at the Stanford AIMI Symposium 2023! Almanac is a retrieval-augmented LLM that provides up-to-date and verifiable answers to medical queries. Link: We benchmark our approach on a novel dataset of clinical scenarios (n = 130) evaluated by a panel of 5 board-certified & resident physicians, and demonstrate significant increases in factuality (mean of 18% at p-value < 0.05) across all specialties. More interestingly, because the retrieved data acts as a single source of truth, we find retrieval-based LLMs to be more robust to prompt injection and manipulation! Future work will involve expanding the scope of our dataset to more specialties and multimodal settings. #Medtwitter #MedEd

Cyril Zakka, MD

18,128 görüntüleme • 3 yıl önce

Stop believing the lie that we have to build more data centers to compete with China. The US already has 5,381 data centers, which is 12 times as many as China’s 449. US billionaires want more data centers to replace our jobs and the US government wants them to surveil us at home and wage more endless wars abroad. The American people need to wake up and understand that while China is investing in bettering life for its people with housing, education, healthcare, and infrastructure, our own government is investing in making life worse for people both at home and abroad.

Stop believing the lie that we have to build more data centers to compete with China. The US already has 5,381 data centers, which is 12 times as many as China’s 449. US billionaires want more data centers to replace our jobs and the US government wants them to surveil us at home and wage more endless wars abroad. The American people need to wake up and understand that while China is investing in bettering life for its people with housing, education, healthcare, and infrastructure, our own government is investing in making life worse for people both at home and abroad.

Power to the People ☭🕊

37,974 görüntüleme • 1 ay önce

In the State of Florida, we will not restrict data centers. We’re not going to regulate data centers. Mark my words, we will ban data centers in the Sunshine State.

In the State of Florida, we will not restrict data centers. We’re not going to regulate data centers. Mark my words, we will ban data centers in the Sunshine State.

James Fishback

35,549 görüntüleme • 2 gün önce

𝗛𝗮𝗽𝗽𝘆 𝗠𝗼𝗻𝗱𝗮𝘆, 𝗧𝗲𝗻𝗲𝗼 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆! A new week means new opportunities to unlock the power of social media data! Whether you're running a Teneo node, earning rewards, learning or exploring all we offer, this week is the perfect time to dive deeper into Teneo and make your mark in the decentralized data world. 🌍 #LearnAboutTeneo

𝗛𝗮𝗽𝗽𝘆 𝗠𝗼𝗻𝗱𝗮𝘆, 𝗧𝗲𝗻𝗲𝗼 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆! A new week means new opportunities to unlock the power of social media data! Whether you're running a Teneo node, earning rewards, learning or exploring all we offer, this week is the perfect time to dive deeper into Teneo and make your mark in the decentralized data world. 🌍 #LearnAboutTeneo

Teneo Protocol

117,920 görüntüleme • 1 yıl önce

The global industry standard oracle platform, Chainlink, is now live on the Injective Mainnet. Time to market for dApp developers is critical and now with Injective’s EVM deployment, the iBuild AI dApp creator, onchain financial modules, and Chainlink data streams - developers on Injective can experience one of the fastest time to market anywhere in the industry. Chainlink brings real time Data Streams with sub-second latency to the only blockchain purpose-built for finance. This now gives Injective developers full access to Chainlink’s battle-tested data stream markets across the Injective ecosystem. Helix 🧬 will be the first Injective dApp to integrate Chainlink Data Streams set to power its crypto and RWA markets. Benefits for using Chainlink Data Streams on Injective: ✅ Low latency market data: up to sub-second delivery ✅ Institutional reliability: collaborating with institutional data giants like ICE and FTSE Russell, An LSEG Business ✅ Programmability: custom tailor formats, cadence, and fields to match a dApps exact needs. Injective markets now powered by Chainlink Data Streams. Ninjas are just getting started. 🥷

The global industry standard oracle platform, Chainlink, is now live on the Injective Mainnet. Time to market for dApp developers is critical and now with Injective’s EVM deployment, the iBuild AI dApp creator, onchain financial modules, and Chainlink data streams - developers on Injective can experience one of the fastest time to market anywhere in the industry. Chainlink brings real time Data Streams with sub-second latency to the only blockchain purpose-built for finance. This now gives Injective developers full access to Chainlink’s battle-tested data stream markets across the Injective ecosystem. Helix 🧬 will be the first Injective dApp to integrate Chainlink Data Streams set to power its crypto and RWA markets. Benefits for using Chainlink Data Streams on Injective: ✅ Low latency market data: up to sub-second delivery ✅ Institutional reliability: collaborating with institutional data giants like ICE and FTSE Russell, An LSEG Business ✅ Programmability: custom tailor formats, cadence, and fields to match a dApps exact needs. Injective markets now powered by Chainlink Data Streams. Ninjas are just getting started. 🥷

Injective 🥷

116,813 görüntüleme • 7 ay önce

✨Instructions:✨ ARMY worked together and was able to retrieve the password for Bombie. Website: Password: BombieRocks2026 Next step: Help Bombie unlock all the data fragments and be ready and get ready for BTS Comeback! He lost some data fragments and he needs to wait 48h to unlock everything🙁... But there is another possibility. Army needs to work together to meet the requirement. If you hover over the blocked buttons you will see the requirements. Until now ARMY unlocked: -Data Fragment 01,02,03,04,08 Explore the website and see what other surprises are there 👀 Hint: In order to unlock all data fragments you need to work together! 😉

✨Instructions:✨ ARMY worked together and was able to retrieve the password for Bombie. Website: Password: BombieRocks2026 Next step: Help Bombie unlock all the data fragments and be ready and get ready for BTS Comeback! He lost some data fragments and he needs to wait 48h to unlock everything🙁... But there is another possibility. Army needs to work together to meet the requirement. If you hover over the blocked buttons you will see the requirements. Until now ARMY unlocked: -Data Fragment 01,02,03,04,08 Explore the website and see what other surprises are there 👀 Hint: In order to unlock all data fragments you need to work together! 😉

Nicolle⁷

61,155 görüntüleme • 3 ay önce

Attio x @Linear Capture feedback, ship with context. → Log customer requests to Linear from any company, person, or deal record in Attio → Highlight text in a call transcript to link the customer voice and meeting to an issue Download now:

Attio x @Linear Capture feedback, ship with context. → Log customer requests to Linear from any company, person, or deal record in Attio → Highlight text in a call transcript to link the customer voice and meeting to an issue Download now:

Attio

11,551 görüntüleme • 10 ay önce