NVILA, a VLM, enhances VILA by scaling spatial and temporal resolutions before compressing visual tokens, enabling efficient high-resolution image & long video processing. Cuts training costs by 4.5X, improves memory & latency, and outperforms top VLMs on benchmarks. Code & models will be released 🔜
December 6, 2024 at 6:47 AM
NVILA, a VLM, enhances VILA by scaling spatial and temporal resolutions before compressing visual tokens, enabling efficient high-resolution image & long video processing. Cuts training costs by 4.5X, improves memory & latency, and outperforms top VLMs on benchmarks. Code & models will be released 🔜
December 5, 2024 at 3:24 AM
Liquid AI introduces synthesis of tailored architectures (STAR) a new approach to automate neural network design tailored to various tasks and hardware setups.
🔗: www.liquid.ai/research/aut...
🔗: www.liquid.ai/research/aut...
December 2, 2024 at 11:45 PM
Liquid AI introduces synthesis of tailored architectures (STAR) a new approach to automate neural network design tailored to various tasks and hardware setups.
🔗: www.liquid.ai/research/aut...
🔗: www.liquid.ai/research/aut...
Marconi: Prefix Caching for the Era of Hybrid LLMs
paper: arxiv.org/abs/2411.19379
Marconi improves caching for hybrid LLMs with policies optimizing reuse likelihood and compute savings, achieving 34.4× higher token hit rates and significantly reducing latency.
paper: arxiv.org/abs/2411.19379
Marconi improves caching for hybrid LLMs with policies optimizing reuse likelihood and compute savings, achieving 34.4× higher token hit rates and significantly reducing latency.
December 2, 2024 at 9:35 AM
Marconi: Prefix Caching for the Era of Hybrid LLMs
paper: arxiv.org/abs/2411.19379
Marconi improves caching for hybrid LLMs with policies optimizing reuse likelihood and compute savings, achieving 34.4× higher token hit rates and significantly reducing latency.
paper: arxiv.org/abs/2411.19379
Marconi improves caching for hybrid LLMs with policies optimizing reuse likelihood and compute savings, achieving 34.4× higher token hit rates and significantly reducing latency.
December 2, 2024 at 9:29 AM
Training Agents with Weakly Supervised Feedback from Large Language Models
paper: arxiv.org/abs/2411.19547
paper: arxiv.org/abs/2411.19547
December 2, 2024 at 7:36 AM
Training Agents with Weakly Supervised Feedback from Large Language Models
paper: arxiv.org/abs/2411.19547
paper: arxiv.org/abs/2411.19547
Critical Tokens Matter: Token-Level Contrastive Estimation Enhence LLM's Reasoning Capability
paper: arxiv.org/abs/2411.19943
paper: arxiv.org/abs/2411.19943
December 2, 2024 at 6:11 AM
Critical Tokens Matter: Token-Level Contrastive Estimation Enhence LLM's Reasoning Capability
paper: arxiv.org/abs/2411.19943
paper: arxiv.org/abs/2411.19943
Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision
paper: arxiv.org/abs/2411.16579
project page: mathcritique.github.io
paper: arxiv.org/abs/2411.16579
project page: mathcritique.github.io
November 26, 2024 at 4:32 AM
Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision
paper: arxiv.org/abs/2411.16579
project page: mathcritique.github.io
paper: arxiv.org/abs/2411.16579
project page: mathcritique.github.io
XGrammar: Flexible and Efficient Structured Generation Engine for Large Language Models
paper: arxiv.org/abs/2411.15100
paper: arxiv.org/abs/2411.15100
November 25, 2024 at 5:19 AM
XGrammar: Flexible and Efficient Structured Generation Engine for Large Language Models
paper: arxiv.org/abs/2411.15100
paper: arxiv.org/abs/2411.15100
August 29, 2023 at 6:38 PM
Crazy how many “web or web3 people “ don’t know anything about IPFS.
August 10, 2023 at 3:12 PM
Crazy how many “web or web3 people “ don’t know anything about IPFS.
Me scrolling the what’s hot feed
April 25, 2023 at 7:08 PM
Me scrolling the what’s hot feed
“Thanks for the bluesky invite, now what do I do?”
April 24, 2023 at 12:27 AM
“Thanks for the bluesky invite, now what do I do?”
40||||
this piece took a month to finish & 150+ layers to get the lighting just right.
You’ll notice a lot of my work has a 🟦 background that’s bc I create on blue and use yellow white, & orange to draw out my design/details. 🫠
this piece took a month to finish & 150+ layers to get the lighting just right.
You’ll notice a lot of my work has a 🟦 background that’s bc I create on blue and use yellow white, & orange to draw out my design/details. 🫠
March 9, 2023 at 1:22 AM
40||||
this piece took a month to finish & 150+ layers to get the lighting just right.
You’ll notice a lot of my work has a 🟦 background that’s bc I create on blue and use yellow white, & orange to draw out my design/details. 🫠
this piece took a month to finish & 150+ layers to get the lighting just right.
You’ll notice a lot of my work has a 🟦 background that’s bc I create on blue and use yellow white, & orange to draw out my design/details. 🫠
