全部/知识/本周高分

LessWrong · 本周高分

5分钟前更新

HISTORY近 30 天历史柱高表示当天去重热搜数量

6017 6718 5619 6120 5821 6022 6223 6424 6225 6226 5727 5728 6529 6130 6801 6102 6203 5504 5605 6006 6407 6408 6609 5810 5611 5812 6113 5814 6315 6216

09/17—10/16 有历史数据

01364
A critical look at Lean4
评论：40 · 标签：Formal Proof、Lean (proof assistant)、AIMilo Moses
02246
OpenAI Can't Self-Police Because It Fires the People Raising the Alarm
评论：3 · 标签：AITurnTrout
03225
Labs’ current alignment plans seem terrible
评论：48 · 标签：AIDaniel Tan
04200
Research Lessons I Learned at MATS & Redwood
评论：16 · 标签：MATS Program、Redwood Research、Rationality、AI、PracticalJo Jiao
05187
Linkpost: Neural Networks Are Already Clean
评论：21 · 标签：Interpretability (ML & AI)、AIYair Halberstadt
06180
Stop telling people to pivot to AI safety, I beg you
评论：58 · 标签：AIHaoxing Du
07160
Quantity vs quality in pottery is a lie
评论：17 · 标签：Epistemic Spot Check、Fact posts、World Modelingbecausecurious
08133
Personas in a Pre-Internet Language Model
评论：6 · 标签：AIDavid Africa
09125
Not Deciphering The Voynich Manuscript
评论：32 · 标签：World Modelingrba
10122
Your prompt… plus everything you've ever (publicly) written
评论：13 · 标签：AIkromem
11111
More people should write basic AI safety explainers
评论：20 · 标签：AILinch
12107
GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior
评论：14 · 标签：AIAxel Højmark
13106
I can't believe it's not BUTTER!
评论：16 · 标签：Deception、Humor、Law and Legal systems、Marketing、World Modelingbecausecurious
14104
An Alignment Forum for AIs? (or: Verification in the Age of Slop)
评论：27 · 标签：AIRaemon
15102
Minimum required pacing: finish aligning it before you train the next one
评论：13 · 标签：Slowing Down AI、AIorthonormal
1699
A short review of "The Zizians: Inside a Trans Death Cult"
评论：10 · 标签：Dramajessicata
1794
Hello World, AI Doompop
评论：13 · 标签：AIalkjash
1890
My Buddy Mike, Efficiency Gains From AI, and The World’s Most Boringly Benign Criminal Conspiracy
评论：7 · 标签：Bureaucracy、Moral Mazes、Organizational Culture & Designabstractapplic
1986
Base Models Can Detect Hidden Intentions that Prompted Classifiers Miss
评论：6 · 标签：LLM Personas、Subliminal Learning、AIOz Amram
2082
A Chinese Perspective on “On Social Reality in China”
评论：2 · 标签：China、World ModelingDusty
2181
Some interesting internal documents from Anthropic litigation
评论：10 · 标签：AIloops
2274
Technical people: actually consider comms
评论：1 · 标签：AImeemi
2373
Arguments About AI Consciousness Remain Highly Motivated and At Best Overconfident
评论：3 · 标签：AIZvi
2469
I trained GPT-2 with full-batch gradient descent so you don't have to
评论：7 · 标签：AIshawnghu
2559
Preface to the Computational Agent Theory and ASI Alignment Sequence
评论：0 · 标签：Agent Foundations、Research Agendas、AIVanessa Kosoy
2656
Game Design Creativity evals
评论：17Raemon
2748
The Continual Constitutional Convention: How language models revise the documents that govern them
评论：0 · 标签：Value Drift、AIpeterkirgis
2847
OpenAI Fires Three Safety Researchers
评论：0 · 标签：AIZvi
2946
[Paper] Distillation for Incrimination and Distillation for Capabilities
评论：2 · 标签：AIsebastian_prasanna
3046
Towards a Better Theory of Anthropics: Introducing the PPA
评论：12 · 标签：Anthropics、World ModelingOnid
3146
The Safety Culture at the Labs is Broken
评论：2 · 标签：AIZvi
3245
The basic argument for making AI better at moral philosophy
评论：1 · 标签：AICaspar Oesterheld crossposts
3343
New Math from OpenAI
评论：5 · 标签：AIZvi
3441
AGI as an Infinite Digital Spain
评论：2 · 标签：AIBearly Legible
3540
Sleep puzzle system
评论：8 · 标签：PracticalKatjaGrace
3639
(Why) we’re launching the next cohort of Generator
评论：0 · 标签：AI Alignment Fieldbuilding、AIAris
3735
An update on Generality Labs
评论：9 · 标签：AIJustin Olive
3834
Astra and Sol can take low probability actions with no CoT
评论：0 · 标签：AI Evaluations、Chain-of-Thought Alignment、AIZephaniah Roe
3933
The user in the LLM might be a moral patient too (a strange implication of the PSM and simulators)
评论：7 · 标签：Interpretability (ML & AI)、AI、World Modelingagastyasridharan
4030
More research needed?
评论：5 · 标签：AIdjbinder
4129
AI #190: Scion of the Times
评论：1 · 标签：AIZvi
4229
Deadlock in the Parliament of the Self
评论：3 · 标签：Shard Theory、Rationality、World ModelingLorxus
4328
Anthropic Has Told The World Little About Their Rogue Claudes
评论：0 · 标签：Anthropic (org)、AIniplav
4428
Filling in the convex hull
评论：2 · 标签：Creativity、Language Models (LLMs)、Logic & Mathematics、AIgjm
4528
The Expensive-Loop Problem
评论：2 · 标签：AIfrmsaul
4627
Manifolds for Five-Year Olds
评论：0 · 标签：AIaxel_sdq
4726
AI Scaling vs Human Scaling
评论：0 · 标签：AIToby_Ord
4826
'Employee-like' ≠ Employee-level
评论：3 · 标签：AIkeltan
4925
Research Priorities for Multi-Agent Swarms
评论：0 · 标签：AIPepijn Cobben
5025
China AI Safety Regulation Monthly Monitor, Issue 1 (September 2026)
评论：0 · 标签：AI Governance、China、Law and Legal systems、Open Source AI、Regulation and AI Risk、AISofia Yablonskaya