when we finally get to see o3's reasoning traces:
LLMS
-
Simplifying Transformer Blocks: Key Research Paper
By
–
Yes. There was a nice paper on that, which looked into some other modifications: Simplifying Transformer Blocks, https://
arxiv.org/abs/2311.01906 -
Merging Query and Key Weight Matrices in LLM Training
By
–
Can we merge the query and key weight matrices in an LLM into a single covariance matrix and still train effectively? Here are some promising early results from a reader: https://
github.com/rasbt/LLMs-fro
m-scratch/discussions/517
…
Anyone else familiar with projects that tried this? -
Chinese Characters Information Density DeepSeek Training Efficiency
By
–
Did the higher information density of Chinese characters contribute to the training efficiency of #DeepSeek?
#AI https://
news.google.com/read/CBMivgFBV
V95cUxORENUcG01RXRtZHAyYUo5UHNvTWVqeGFKZFVqc01rWEhfUHMzWXlGaGtFRk1MVzVmbHRnenlBWTVRUVVCanNyREV0Z29BeE1aaFJnaFVPSUhWUnBVQ2V1MmxTT0Nqa0pPOWhqUHhjd1A2QlpIei1lai1FNXJwZDlDc3NUVnNNTHNic19qVUp5MXJjOTRaLUtBWmk4UXY5c0YzbEdTdkpBVnVkNVp4dVREX0hOZDliX2Q1Z09NeWpB0gG-AUFVX3lxTE5VUmNUTm5lRldkaUZaZE9RaW5STDhneUhQdzFZeDFtcXFBU24xY2RTWWwxR1l2R2JqclpHYWVEWlljTHFRRm41OVFWbXJYLVlubzJsWmxCeTludWFuOGNienltUW1tS0hDbkkzYWJ0anAtZHlXOEZ0VVY0eUF6anJhUjVaR1RRRG9RTTVkUTBsQWpNYzVEeTR4dExCN25hV3E0X3AwWGczMnBNSFEySTd6dXJGZTVzY2piZHlGc3c?hl=en-US&gl=US&ceid=USen
… -
A complete cheatsheet for mastering DeepSeek AI
By
–
DeepSeek is the most powerful AI tool right now. But most people are using it wrong. Here’s the complete cheatsheet to master DeepSeek easily:
-
OpenAI Releases Reasoning Best Practices Guide for o1 and o3-mini
By
–
It’s been awesome to watch the adoption of our o1 and o3‑mini models. We’ve just published a new guide on reasoning best practices, with real-world tips from some of the best AI founders using them in production. What else can we ship to help you build?
-
Mistral Small 3 Now Available on Poe Platform
By
–
Mistral Small 3 is available at https://
poe.com/Mistral-Small-3 and on all Poe apps. (2/2) -
Mistral Small 3 Available on Poe: Fast and Knowledgeable
By
–
Now on Poe: Mistral Small 3! This latest model from Mistral AI is fast (150 tokens/second), knowledgeable (81% on MMLU benchmark), and excels at understanding and following instructions. (1/2)
-
Llamapalooza roadshow comes to Seattle with Meta and AWS
By
–
🦙 🛣️ Llamapalooza is going on the road!
— Cerebras (@cerebras) 14 février 2025
On February 27th – we will be in Seattle for an unforgettable evening of exploring llama models in production, featuring headliners from @AIatMeta, @awscloud, and Cerebras.
Shoutout to our cohosts @ollama and @AITinkerers!
Tickets are… pic.twitter.com/39bR2xngQ3Llamapalooza is going on the road! On February 27th – we will be in Seattle for an unforgettable evening of exploring llama models in production, featuring headliners from
@AIatMeta
, @awscloud
, and Cerebras. Shoutout to our cohosts @ollama and @AITinkerers
!
Tickets are -
DeepSeek R1 Now Available on SambaNova Cloud Platform
By
–
The BIG WHALE doing GOAT things #DeepSeekR1 is running on SambaNova Cloud now! Whale you try it? http://
cloud.sambanova.ai