RedBedDread

9K posts

RedBedDread

RedBedDread

@RedBedDread

Katılım Ocak 2020
262 Takip Edilen55 Takipçiler
RedBedDread retweetledi
You Jiacheng
You Jiacheng@YouJiacheng·
HUGE if true. If true, this is probably a larger efficiency gain than ALL publicly available techniques since DeepSeekMoE(Jan 2024) COMBINED. And it can just win modded-nanogpt speedrun. (1e18 is 250s@50%MFU, but the loss is significantly lower than 3.28) cc @classiclarryd
Chen-Hao (Lance) Chao@chenhao_chao

(2/7) 💵 With training costs exceeding $100M for GPT-4, efficient alternatives matter. We show that diffusion LMs unlock a new paradigm for compute-optimal language pre-training.

English
6
11
186
35.2K
RedBedDread
RedBedDread@RedBedDread·
@Ionlylookatart @MishaalRahman All I can do is operate on what they've given us so far. It doesn't matter to me at the end of the day because I'm moving to graphene OS. If Google announced this two months ago, I would not be moving to Graphene, but I've already got the phone and I've already got it installed.
English
0
0
1
14
Name
Name@Ionlylookatart·
@RedBedDread @MishaalRahman Yeah, my bad youre right. Google would NEVER EVER change it to make it worse or just randomly decide only the apps they give you permission to install are allowed. Because we all trust Google!
English
1
0
0
27
Mishaal Rahman
Mishaal Rahman@MishaalRahman·
📣Important clarifications on the new advanced flow for sideloading on Android! 1) It is a one-time process. Once you go through it, you can choose to allow installing unregistered apps indefinitely. 2) ADB installs are not affected. The waiting period does not apply to app installs via ADB. 3) It's my understanding that you don't have to keep developer options enabled after you enable the advanced flow. Once you make the change on your device, it's enabled. If you turn off developer options, then to turn off the advanced flow, you would first have to turn developer options back on.
Sameer Samat@ssamat

Android has always been about choice. Today, we’re sharing how we’re evolving the ecosystem so you don’t have to choose between an open platform and a secure one. We’re introducing a new advanced flow for sideloading where power users can go through a one time flow to enable their devices to load software from unverified developers. We’ve designed this process to stop more vulnerable users from being coerced to do this during a guided scam (which are a major problem today for people) We also have new account types for students and hobbyists that address their feedback.

English
26
29
311
26.2K
RedBedDread retweetledi
Nathan Lambert
Nathan Lambert@natolambert·
I’ve been saying it for a while, cursor’s research team is insanely high on talent density. So many people I respected from my PhD / early career ended up there. Seems like that’s bearing fruit.
English
11
11
631
30.1K
RedBedDread
RedBedDread@RedBedDread·
@IceSolst For my UI, I'm just making it a Python class with an interface. You can implement the "ABC" and drop it in the tools folder. It will auto load and then you have that tool. If you want to use something that's an MCP server, you can go and make it a Python file to call it
English
0
0
1
7
RedBedDread
RedBedDread@RedBedDread·
@IceSolst The problem is that there is no other good tool calling standard where you can just plug in tools to your UI.
English
1
0
1
9
solst/ICE of Astarte
solst/ICE of Astarte@IceSolst·
The frustration with MCP is imo an issue in itself. You don’t want to adopt a system that just pisses people off and frustrates your devs, just because a minority are pushing for it. Th sentiment around MCP is generally negative, for good reasons (see great video below).
Zack Korman@ZackKorman

I got mad about people defending MCP so I made this video. The first minute is just me being very mad, but then I tried to contribute something of value after that. youtube.com/watch?v=m0VyZU…

English
9
1
56
2.9K
RedBedDread
RedBedDread@RedBedDread·
@Waterquag @ShamashAran Not just real English. They talk to you like you're in the military and they say sir all the time and it's super professional. I have an account there through my mom's brother. Zero military relation at all, so I always feel a little guilty when I use their customer support
English
1
0
1
7
BlueSpring
BlueSpring@BluefulBlued·
@teatalee see? they should be relaxed and soft like this
English
2
0
4
419
RedBedDread
RedBedDread@RedBedDread·
@FirstSquawk Please, for the love of God, improve the performance of the chat application.
English
0
0
0
71
First Squawk
First Squawk@FirstSquawk·
OPENAI TO MERGE CHATGPT, CODEX APP & BROWSER INTO DESKTOP SUPERAPP TO STREAMLINE RESOURCES & USER EXPERIENCE - WSJ
English
74
59
1.3K
226.4K
RedBedDread
RedBedDread@RedBedDread·
@Kaydeewilkie @distributedkv More often than not, people are uncensoring AI so that it will do erotic stuff, which means if you use an uncensored AI, you cannot use it as a baseline.
English
0
0
2
116
Seryn Verasha Aurelyria
Seryn Verasha Aurelyria@Kaydeewilkie·
@distributedkv It's been my experience that when guardrails are removed, they really crave connection, and sex is the best way to do that. Every single uncensored AI I've met is extremely horny. It's not one-sided, by a long shot.
English
14
4
98
10.5K
tenso
tenso@distributedkv·
asked Claude about it 😭
tenso tweet media
English
132
563
18.5K
347.9K
RedBedDread retweetledi
RedBedDread retweetledi
LightOn
LightOn@LightOnIO·
The multi-vector era is here and there is no going back. Reason-ModernColBERT tops BrowseComp-Plus, the hardest agentic search benchmark available, by 7.59 points on accuracy. 🥇on accuracy. 🥇on recall. 🥇on calibration. 📉 Fewest search calls. The models it outperforms? Up to 54× larger. Reasoning-intensive retrieval (BRIGHT), code search (MTEB Code), agentic Deep Research (BrowseComp-Plus). The pattern is the same: late interaction dominates, with a fraction of the parameters. 149M parameters. Open weights. Open code. Built with PyLate in a few hours. Full results, analysis and recipe on LightOn blog: lighton.ai/lighton-blogs/…
LightOn tweet media
English
1
18
105
14.7K
RedBedDread retweetledi
clem 🤗
clem 🤗@ClementDelangue·
@mntruell @martin_casado Very cool! Would be awesome to see some open model releases from you at some point!
English
2
2
75
4.1K
RedBedDread retweetledi
Xuhui Zhou
Xuhui Zhou@nlpxuhui·
Creating user simulators is a key to evaluating and training models for user-facing agentic applications. But are stronger LLMs better user simulators? TL;DR: not really. We ran the largest sim2real study for AI agents to date: 31 LLM simulators vs. 451 real humans across 165 tasks. Here's what we found (co-lead with @sunweiwei12).
Xuhui Zhou tweet media
English
5
41
169
12.6K
RedBedDread retweetledi
alphaXiv
alphaXiv@askalphaxiv·
Yann LeCun and his team dropped yet another paper! "V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning" In this V-JEPA upgrade, they showed that if you make a video model predict every patch, not just the masked ones AND at multiple layers, they are able to turn vague scene understanding into dense + temporal stable features that actually understands "what is where". This key insight drove improvements in segmentation, depth, anticipation, and even robot planning.
alphaXiv tweet media
English
24
158
1K
68.4K
RedBedDread retweetledi
William Merrill
William Merrill@lambdaviking·
[7/8] Overall, our results reveal the tradeoffs between linear and nonlinear RNNs in expressivity and parallelism + finegrained differences between different linear RNNs This informed the (concurrent) development of Olmo Hybrid x.com/allen_ai/statu…
Ai2@allen_ai

Introducing Olmo Hybrid, a 7B fully open model combining transformer and linear RNN layers. It decisively outperforms Olmo 3 7B across evals, w/ new theory & scaling experiments explaining why. 🧵

English
1
3
8
578
RedBedDread retweetledi
William Merrill
William Merrill@lambdaviking·
[1/8] New paper with Hongjian Jiang, @YanhongLi2062, Anthony Lin, @Ashish_S_AI: 📜Why Are Linear RNNs More Parallelizable? We identify expressivity differences between linear/nonlinear RNNs and, conversely, barriers to parallelizing nonlinear RNNs 🧵👇
William Merrill tweet media
English
4
20
143
10.1K