MATS Research

133 posts

MATS Research banner
MATS Research

MATS Research

@MATSprogram

MATS empowers researchers to advance AI alignment, transparency, and security

Berkeley, CA 가입일 Kasım 2023
136 팔로잉3.5K 팔로워
고정된 트윗
MATS Research
MATS Research@MATSprogram·
1/ 🚨 MATS Autumn 2026 applications are now open. 10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders. 📍 Berkeley + London 📅 Sep 28 – Dec 4, 2026 💰 $5000/month stipend + $8,000/month compute Apply by June 7 AoE ↓
English
10
93
731
3.2M
MATS Research 리트윗함
Alex Turner
Alex Turner@Turn_Trout·
MATS Autumn applications due June 7! Pitch: Come work with me and Alex Cloud in Team Shard! We have fun, consistently make real alignment progress (we pioneered steering vectors in 2023!), and help scholars tap into their latent abilities.
Alex Turner tweet media
English
3
9
148
9.6K
MATS Research
MATS Research@MATSprogram·
The essentials: 📅 10 weeks · Sep 28 – Dec 4, 2026 💰 $5k/mo stipend + $8k/mo compute & research budget 📍 Berkeley / London / remote 🏠 Housing, meals, travel, J-1 visa ↗ 6–12 month extension opportunity
English
1
0
10
1.3K
MATS Research
MATS Research@MATSprogram·
🚨 New for MATS Autumn 2026: the Founding & Field-Building track. A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives. Apply by June 7 AoE ↓
English
2
20
237
27.9K
MATS Research 리트윗함
Dylan Feng
Dylan Feng@dylanfeng_·
🚨 New research work with @CHAI_Berkeley! We provide the first multi-domain benchmark evaluating safety monitors for OOD misalignment detection by intentionally restricting the training dataset. Special thanks to folks @MATSprogram and @haizelabs for providing valuable feedback and compute.
Cassidy Laidlaw@cassidy_laidlaw

We've seen AI models deceive, gaslight, and drive users to psychosis—safety issues that labs didn't anticipate until they caused real harm. We built the first benchmark of these unknown unknown alignment failures and found that OOD detection can help prevent them. 🧵

English
0
2
7
943
MATS Research 리트윗함
jasmine
jasmine@jasminexli·
My @MATSprogram research is out, go read! @turntrout and I propose eval cooperativeness -- a target disposition with early results on making eval-aware models behave more consistently. Some personal reflections on my research & MATS:🧵
Alex Turner@Turn_Trout

New research from Team Shard & @jasminexli! AIs increasingly fake good behavior, which might ruin our ability to evaluate models. We trained models to be 𝘦𝘷𝘢𝘭-𝘤𝘰𝘰𝘱𝘦𝘳𝘢𝘵𝘪𝘷𝘦: to want to give evaluators accurate info. This surfaces hidden misalignment! 🧵

English
6
9
117
9.7K
MATS Research 리트윗함
Luke Drago
Luke Drago@luke_drago_·
If you want to research interventions to gradual disempowerment or the intelligence curse, @LRudL_ and I are mentoring a @MATSprogram stream this autumn. Many people have asked me “what’s the plan to make this go well?” Right now, there’s not one. You should help fix that. 🧵
English
4
19
183
18.6K
MATS Research 리트윗함
Alex Turner
Alex Turner@Turn_Trout·
New research from Team Shard & @jasminexli! AIs increasingly fake good behavior, which might ruin our ability to evaluate models. We trained models to be 𝘦𝘷𝘢𝘭-𝘤𝘰𝘰𝘱𝘦𝘳𝘢𝘵𝘪𝘷𝘦: to want to give evaluators accurate info. This surfaces hidden misalignment! 🧵
Alex Turner tweet media
English
4
19
122
13.4K
MATS Research 리트윗함
Julian Minder
Julian Minder@jkminder·
New blog! Synthetic Persona Pretraining (SPP): Alignment from Token Zero Current alignment is shallow - values bolted on after pretraining can be routed around. To solve this, we wrote the desired persona directly into pretraining data. Early results, but we're very excited. 🧵
Julian Minder tweet media
English
17
39
300
44.9K
MATS Research 리트윗함
Ryan Kidd
Ryan Kidd@ryan_kidd44·
MATS Autumn 2026 applications are live! We also have new Founding & Fieldbuilding and Biosecurity tracks! Come build the AI safety & security institutions of the future!
MATS Research@MATSprogram

1/ 🚨 MATS Autumn 2026 applications are now open. 10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders. 📍 Berkeley + London 📅 Sep 28 – Dec 4, 2026 💰 $5000/month stipend + $8,000/month compute Apply by June 7 AoE ↓

English
8
23
1.1K
3.1M
Choudhary
Choudhary@PotatoChoudhary·
@MATSprogram hay, i don't really have anyone for a LOR, should i still apply ?
English
1
0
1
2.2K
MATS Research
MATS Research@MATSprogram·
1/ 🚨 MATS Autumn 2026 applications are now open. 10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders. 📍 Berkeley + London 📅 Sep 28 – Dec 4, 2026 💰 $5000/month stipend + $8,000/month compute Apply by June 7 AoE ↓
English
10
93
731
3.2M
MATS Research
MATS Research@MATSprogram·
@akhil_manga Our Founding & Field-Building track might be similar to Generator, but potentially more focused on founding.
English
1
0
2
473
MATS Research
MATS Research@MATSprogram·
@sting_punk We allow remote participation and can help with visas in the US and UK!
English
0
0
0
15
MATS Research 리트윗함
Oscar Gilg
Oscar Gilg@gilg_oscar·
First preprint! Working with @patrickbutlin during @MATSprogram. LLM Assistant personas like being helpful, evil personas like being harmful. We found that a single direction represents helping as good under the Assistant, and ‘harm’ as good under evil.
Oscar Gilg tweet media
English
5
18
94
11.6K
MATS Research
MATS Research@MATSprogram·
6/ Reducing risks from powerful AI is one of the world's most urgent and talent-constrained challenges. Know someone who'd be a strong fit? Share this thread. 🔗 matsprogram.org/apply?utm_sour…🚨 June 7, 2026 AoE
English
0
0
47
5.3K

탐색

@CHAI_Berkeley @haizelabs @Turntrout @LRudL_ @jasminexli @veroskuro @akhil_manga @sting_punk