User contributions for Catai

4 April 2025

• 21:4921:49, 4 April 2025 diff hist +1,548‎ stat946W25 ‎ →‎Masked-Diffusion LM: Faster and Smarter: : Added more details and examples to Masked-Diffusion LM.

29 March 2025

• 16:1216:12, 29 March 2025 diff hist +3,737‎ stat946W25 ‎ →‎Topic 7: Dynamic Models: Many-in-One Language Models: : Added a description for SHARCS.

22 March 2025

• 22:1922:19, 22 March 2025 diff hist +1,311‎ stat946W25 ‎ →‎Topic 4: Quantization: : Added more to the beginning to provide an overview of the technologies before more description is given on each one

12 March 2025

• 22:0422:04, 12 March 2025 diff hist +21‎ stat946W25 ‎ →‎Flash Attention V3: : Fixed formatting for better readability
• 22:0222:02, 12 March 2025 diff hist +1,578‎ stat946W25 ‎ →‎Topic 11: Flash Attention: : Added section for Flash Attention V3 and elaborated on the technique
• 21:3521:35, 12 March 2025 diff hist +178‎ stat946W25 ‎ →‎Flash Attention V1: : Add the key point of Flash Attention V1 and transition starter to the detailed description.

9 March 2025

• 15:2815:28, 9 March 2025 diff hist +4‎ stat946W25 ‎ →‎SpAtten: : Adjusted spacing for readability
• 15:2715:27, 9 March 2025 diff hist +2,322‎ stat946W25 ‎ →‎SpAtten: : Elaborated on the technique with more details
• 14:5614:56, 9 March 2025 diff hist 0‎ stat946W25 ‎ →‎Attention with Linear Biases (ALiBi): : Fixed typo.
• 14:5514:55, 9 March 2025 diff hist −71‎ stat946W25 ‎ →‎Sparse Sinkhorn Attention: : Reword the sorting section to be block specific given its heading.
• 14:5114:51, 9 March 2025 diff hist +253‎ stat946W25 ‎ →‎Sparse Sinkhorn Attention: : Added more description to the sorting algorithm and fixed typos.
• 12:2012:20, 9 March 2025 diff hist +1,091‎ stat946W25 ‎ →‎Topic 12: State Space Models: : Added key takeaway section with table to summarize the main wiki points
• 11:3811:38, 9 March 2025 diff hist −15‎ stat946W25 ‎ →‎Diagonal State Space Model (DSS): : Edited the transition sentence for better wiki flow.
• 11:3711:37, 9 March 2025 diff hist +975‎ stat946W25 ‎ →‎Structured State Space (S4): : Provided more description on the model and its shortcomings
• 11:2711:27, 9 March 2025 diff hist +2‎ stat946W25 ‎ Separated equations from description in core concepts section
• 11:2711:27, 9 March 2025 diff hist −1‎ stat946W25 ‎ Formatted spacing for core concepts
• 11:2711:27, 9 March 2025 diff hist +3‎ stat946W25 ‎ Formatted core concepts better with spacing
• 11:2611:26, 9 March 2025 diff hist +229‎ stat946W25 ‎ Added further description to "Core concepts"