mrgam @mrgammm

Introducing Claude Science, a new app designed with every stage of research in mind. Artifacts traced to their code, environments managed on demand, and 60+ optional scientific databases that you can connect. Available now in beta.

1K 3K 35K 9.8M 15K

View Details

mrgam @mrgammm

4 months ago

@liquidai Are you guys gonna add more language like Indonesia next time? This is nice for live translation etc

0 0 0 973 0

View Details

mrgam @mrgammm

4 months ago

@UnslothAI I tried running it on a jetson orin nano with 8gb got at best 52tok/s any idea to adjust? Was hoping to get higher, but hit a bit of a brick wall

0 0 0 20 0

View Details

mrgam @mrgammm

4 months ago

@dani_avila7 This is gonna be fire, rather then having many subscription

0 0 0 76 0

View Details

Ning Yu @realNingYu

4 months ago

📢 @eyelinestudios and @netflix latest research, 🌟Go-with-the-Track🌟, has been accepted to #SIGGRAPH2026. We propose a video generation framework that jointly conditions on multiple reference images and reference-anchored point tracks, enabling a range of content creation applications within a unified model: • Motion-preserving video restylization using multiple reference images; • Mesh- or keypoint-driven compositing and stylization; • Camera motion control for both static and dynamic scenes using multiple reference images captured from different view-points and time steps. ✊Congrats to the team: @Koichi_N_, @yash2kant, Zhizheng Liu, @RyanBurgert, @Yuancheng_Xu0, @kuanhenglin, Emmett Steven, @JulienPhilip2, @maleewahaha, Andrea Vedaldi, @debfx, @realNingYu 📰 Paper: arxiv.org/pdf/2606.20891 🌐 Project: eyeline-labs.github.io/Go-with-the-Tr… ⌨️ Code: github.com/Eyeline-Labs/G… *** This work is part of the ongoing research and development at @eyelinestudios and @netflix, and we look forward to seeing its techniques and workflows adopted in future productions.

3 17 123 29K 92

View Details

mrgam @mrgammm

4 months ago

Honestly the ability to use the /orx locally in claude/codex makes it 10x better

alphaXiv @askalphaxiv

4 months ago

Introducing autoresearch for GitHub repos Change 'Github' to 'ARGithub' in any repo URL Research artifacts extend beyond papers. Autoresearch is especially useful for experimenting on existing codebases that move fast and outpace their own publications. With one URL change you

19 83 697 74K 700

0 0 1 55 0

View Details

Zhengyang Geng @ZhengyangGeng

4 months ago

This is a paper with very good science! Liked it for many reasons. A useful way to think about diffusion: first it decides what kind of sample to generate, only later which sample it converges to. The first transition is spectral; the second is entropic. The interval between them is where the "within-class" generation happens. Perfectly fitting the empirical score eventually leads to memorization (but rarely happens in reality!) Also a nice reminder that simple physical models are far from obsolete. Gaussian mixtures are useful not because the world is Gaussian, but because a good toy model acts as a microscope for mechanisms.

Qfunction@Stat & Bioinfo @StudyQC

4 months ago

nature.com/articles/s4146… これおもろいな。Diffusion Modelの逆拡散過程、は最初ランダムに動いたあとに構造が出現して最後はアトラクタに収束する３段階からなる。しかもその相転移が起こる時間ｔを方程式の解として与えられる

4 185 1K 108K 923

2 31 176 17K 185

View Details

Haozhe Jiang @erichzjiang

4 months ago

Why aren’t Diffusion Language Model smart yet? Lacking stable post training is a major bottleneck! Meet DiPOD: the tripod for diffusion model post-training. DiPOD boosts accuracy across reasoning tasks, with Sudoku jumping from 22% to 97%, through a one-line code change. 🧵1/5

13 74 592 140K 472

View Details

mrgam @mrgammm

4 months ago

@ClaudeDevs @bcherny Main problem is the website, as for claude code, does not seem to have a problem. but the website since alot of chatting and giving files it seems to just finish my whole limit and not even compacting i guess?

0 0 0 40 0

View Details

mrgam @mrgammm

4 months ago

@ClaudeDevs Hi guys, could you check the claude.ai web or desktop? seems like auto compaction is weird right now, my account limits the last 2 days got from 30% to now 92% and I haven't really been using it much. Please need help using the max 5x, not like usual cc @bcherny

1 0 1 106 0

View Details

Hila Manor @hila8manor

4 months ago

"Spanning the Visual Analogy Space with a Weight Basis of LoRAs" is heading to #ECCV2026🥳 We learn to dynamically mix LoRAs to improve analogy-based image editing. 🔗research.nvidia.com/labs/par/lorwe… w/ the amazing @RinonGal @HaggaiMaron @GalChechik at @nvidia & @t_michaeli🤩 @eccvconf

4 12 64 4K 21

View Details

mrgam @mrgammm

4 months ago

@xy9301 Congrats! more to go hahaha

0 0 0 54 0

View Details

mrgam @mrgammm

4 months ago

This autoarxiv is interesting

alphaXiv @askalphaxiv

4 months ago

Introducing GLM 5.2 for autoresearch GLM 5.2 is the first open weights model we've tried on our autoresearch pipeline that's proven capable for real research tasks. With Fable 5's restrictions on research, having an open weights alternative is a huge win for open source Watch

26 226 2K 272K 1K

0 0 0 54 0

View Details

Keisuke Okumura @_kei18

4 months ago

最近のお気に入りスライド。D4ormの紹介で使う arxiv.org/abs/2503.12204

0 18 115 10K 83

View Details

mrgam @mrgammm

4 months ago

@workwithdhanes Well noted, Thank you infonya kak! Coba saya ikut join juga

0 0 0 38 0

View Details

Underfox @Underfox3

4 months ago

In this paper is presented linear FDSNet, a neural network for performing the inverse operation of a Hierarchical Off-Diagonal Low-Rank matrix, inspired by a fast direct solver. arxiv.org/pdf/2606.19895