#qualitycompression search results

Ashutosh Maheshwari

Oct 13

Compression techniques I’d study if I wanted small but smart LLMs. Bookmark this. 1.Quantization 2.Distillation 3.Low-Rank Adaptation 4.Weight Sharing 5.Sparse Matrices 6.Layer Dropping 7.Knowledge Transfer 8.Embedding Compression 9.Mixed Sparsity 10. Progressive Shrinking…

asmah2107's tweet image. Compression techniques I’d study if I wanted small but smart LLMs.

Bookmark this.

1.Quantization
2.Distillation
3.Low-Rank Adaptation
4.Weight Sharing
5.Sparse Matrices
6.Layer Dropping
7.Knowledge Transfer
8.Embedding Compression
9.Mixed Sparsity
10. Progressive Shrinking…

BRIAN CALÍREY

@BRIANCALIREY

Nov 4

EQ & Compression Ain’t Saving Your Mix: The Secret You’re Missing

Guri Singh

@heygurisingh

4 h

🚨 DeepSeek just did something wild. They built an OCR system that compresses long text into vision tokens literally turning paragraphs into pixels. Their model, DeepSeek-OCR, achieves 97% decoding precision at 10× compression and still manages 60% accuracy even at 20×. That…

heygurisingh's tweet image. 🚨 DeepSeek just did something wild.

They built an OCR system that compresses long text into vision tokens literally turning paragraphs into pixels.

Their model, DeepSeek-OCR, achieves 97% decoding precision at 10× compression and still manages 60% accuracy even at 20×. That…

The Music Guy

@MusicGuyMixing

1 h

Today I identify what is New York compression and more importantly how and why you should use it to add energy, sustain, and life to your mix. musicguymixing.com/new-york-compr…

MusicGuyMixing's tweet image. Today I identify what is New York compression and more importantly how and why you should use it to add energy, sustain, and life to your mix. musicguymixing.com/new-york-compr…

AK

@_akhaliq

Oct 20

DeepSeek-OCR Contexts Optical Compression

Qihang Zhang@EMNLP 2025

@QihangZhang00

Nov 5

Introducing our work Test-Time Steering for Lossless Text Compression via Weighted Product of Experts — a simple way to combine LLMs with traditional compressors so the ensemble is never worse than the best expert, and often better. I will be at Hall C this afternoon! (0/N)

QihangZhang00's tweet image. Introducing our work Test-Time Steering for Lossless Text Compression via Weighted Product of Experts — a simple way to combine LLMs with traditional compressors so the ensemble is never worse than the best expert, and often better.

I will be at Hall C this afternoon! (0/N)

Matthews Technologies

@MatthewsTechno1

Nov 6

🚫 Common Mistake: Uploading high-res images directly from your phone. Every 5MB image slows your page — and costs you conversions. 💡 Use compression tools (TinyPNG, Squoosh, etc.) to shrink file size without losing quality. Fast site = happy user = more sales.

MatthewsTechno1's tweet image. 🚫 Common Mistake:

Uploading high-res images directly from your phone.

Every 5MB image slows your page — and costs you conversions.

💡 Use compression tools (TinyPNG, Squoosh, etc.) to shrink file size without losing quality.

Fast site = happy user = more sales.

SIGKITTEN

@SIGKITTEN

Jul 30

Strap in guys Claude said it not only implemented lossless compression but also somehow made it 2x faster. This is going to be great

SIGKITTEN's tweet image. Strap in guys
Claude said it not only implemented lossless compression but also somehow made it 2x faster.
This is going to be great

@bluecow 🐮(schizo)

@BLUECOW009

Jan 18

compressed a file from 295KB to 328 bytes: Compression Performance: - Original image: 295,468 bytes - Compressed file: 328 bytes - Ratio: 900.82:1 (meaning the compressed file is 900.82 times smaller) - Space savings: 99.89% - Only 0.0089 bits needed per original byte

BLUECOW009's tweet image. compressed a file from 295KB to 328 bytes:

Compression Performance:

- Original image: 295,468 bytes
- Compressed file: 328 bytes
- Ratio: 900.82:1 (meaning the compressed file is 900.82 times smaller)
- Space savings: 99.89%
- Only 0.0089 bits needed per original byte

Alessio Devoto

@devoto_alessio

Oct 2

Expected Attention compresses the KV Cache by estimating the attention score from future queries! Both during prefilling & decoding 🚀 All code released in our KVPress library & Leaderboard!

devoto_alessio's tweet image. Expected Attention compresses the KV Cache by estimating the attention score from future queries! Both during prefilling &amp; decoding 🚀 All code released in our KVPress library &amp; Leaderboard!

quantumBOOM

@quantumboomnow

8 m

$NVDA $GOOG $RGTI $BABA Deeptech 23/50: Optical compression can expand context length by an order of magnitude without extra compute. quantumboomnow.com

vidaio_τ

@vidaio_

Nov 5

Did somebody say 96% compression with no discernible picture quality loss???????!!!!!!!! YES! we did 🚀🚀🚀 Try out our compression model for yourself at vidaio.io

vidaio_'s tweet image. Did somebody say 96% compression with no discernible picture quality loss???????!!!!!!!!

YES! we did 🚀🚀🚀

Try out our compression model for yourself at vidaio.io

Wesley Long 🧱🔺

@weslong

Nov 3

25 years ago I knew this so called freemium web was a con. The Great Compression is the real-time scrubbing of human authorship by AI and tech companies. This is theft of value on a massive scale.

weslong's tweet image. 25 years ago I knew this so called freemium web was a con. The Great Compression is the real-time scrubbing of human authorship by AI and tech companies. This is theft of value on a massive scale.

Chidanand Tripathi

@thetripathi58

Nov 5

Okay... You know that sinking feeling when your image looks stunning in preview, but turns to mush the moment you zoom in? It’s not your fault. Resolution limits ruin great work every day. Let me explain:

thetripathi58's tweet image. Okay...

You know that sinking feeling when your image looks stunning in preview, but turns to mush the moment you zoom in? It’s not your fault.

Resolution limits ruin great work every day.

Let me explain:

Grady

@GradyEditor

Aug 21, 2024

This tool let's you compress 90% of a video - and keeps the same quality. This is a must use for video editors.

Quality Pixels

@Quality_Pixels

Jul 25, 2024

#KeerthySuresh UHD Still!!! Size:2.5MB(2240x3360) Drive: bit.ly/3WbUrk0 @KeerthyOfficial #QualityPixels

Chopper

@chopper__daddy

Jun 20, 2023

I put together this simple tool to help save some gas cost on ethscriptions. This demo shows lossless image compression savings on the file size. From 64kb (~$50) to 9kb (~$7) @ 15gwei This will be available on flooredApe early this week. Features requests? Add them in the…

Harald Schäfer

@___Harald___

May 30, 2024

Interesting to see @ylecun arguing pure ViTs are not practical. We recently moved our compressors over to pure ViTs, no convs. They take longer to train, but work quite well. Left image is compressed to only 224 bytes, right is original.

___Harald___'s tweet image. Interesting to see @ylecun arguing pure ViTs are not practical. We recently moved our compressors over to pure ViTs, no convs.

They take longer to train, but work quite well. Left image is compressed to only 224 bytes, right is original.

Alessio Devoto

@devoto_alessio

Jul 21

🏆 Our @nvidia KV Cache Compression Leaderboard is now live! Compare state-of-the-art compression methods side-by-side with KVPress. See which techniques are leading in efficiency and performance. 🥇 huggingface.co/spaces/nvidia/…

devoto_alessio's tweet image. 🏆 Our @nvidia KV Cache Compression Leaderboard is now live!

Compare state-of-the-art compression methods side-by-side with KVPress. See which techniques are leading in efficiency and performance. 🥇
huggingface.co/spaces/nvidia/…

Greg Lewis

@Greg_GLD

Mar 22, 2024

Video compression for web is hard. I have learnt this stuff just through trial and error. In this thread, I'll show you how to optimise your video for web. 1/8🧵