RSS

Aether

TagLast edit: 24 Jun 2026 18:19 UTC by Joey Yudelson

Aether is an AI safety research organization.

Im­pli­ca­tions of Con­tinual Learn­ing for LLM Agents: Introduction

RohanS, Rauno Arike, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward and Seth Herd
12 Jun 2026 18:36 UTC
50 points
0 comments6 min readLW link

What’s Con­tinual Learn­ing, and Why Might We Ex­pect To See It In Ad­vanced LLM Agents?

RohanS, Rauno Arike, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward and Seth Herd
12 Jun 2026 18:43 UTC
32 points
2 comments17 min readLW link

How con­cerned should we be about As­tra’s re­cur­rent ar­chi­tec­ture?

Rauno Arike2 Sep 2026 18:28 UTC
170 points
28 comments10 min readLW link

Es­ti­mat­ing No-CoT Task-Com­ple­tion Time Hori­zons of Fron­tier AI Models

Anders Cairns Woodruff, Francis Rhys Ward, Dewi Gould, Rauno Arike, Jason R Brown, Jo Jiao, wlanderson, ariana_azarbal, harrymayne, Patrick Leask, Twm Stone, Josh Hills, Ida Caspary, Shubhorup Biswas and Julian Stastny
10 Jun 2026 17:58 UTC
280 points
23 comments4 min readLW link

How might con­tinual learn­ing af­fect safety and al­ign­ment?

Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward and Seth Herd
13 Jun 2026 17:34 UTC
64 points
2 comments16 min readLW link

Ex­tract-and-Eval­u­ate Mon­i­tor­ing Can Sig­nifi­cantly En­hance CoT Mon­i­tor Perfor­mance (Re­search Note)

Rauno Arike, RohanS and Shubhorup Biswas
8 Aug 2025 10:41 UTC
52 points
7 comments10 min readLW link

A List of Re­search Direc­tions in Char­ac­ter Training

Rauno Arike19 Mar 2026 22:58 UTC
49 points
21 comments8 min readLW link

The Most For­bid­den Tech­nique is not always forbidden

Rauno Arike18 Jul 2026 0:59 UTC
84 points
4 comments8 min readLW link

Hid­den Rea­son­ing in LLMs: A Taxonomy

Rauno Arike, RohanS and Shubhorup Biswas
25 Aug 2025 22:43 UTC
81 points
12 comments12 min readLW link

13 Ar­gu­ments About a Tran­si­tion to Neu­ralese AIs

Rauno Arike7 Nov 2025 16:19 UTC
53 points
14 comments10 min readLW link

Ex­plor­ing Re­in­force­ment Learn­ing Effects on Chain-of-Thought Legibility

Julian H, RohanS, Baram Sosis, vedant-badoni and The-Turtle
6 Jan 2026 3:04 UTC
42 points
3 comments21 min readLW link

[Paper] How does in­for­ma­tion ac­cess af­fect LLM mon­i­tors’ abil­ity to de­tect sab­o­tage?

Rauno Arike, Raja Moreno, RohanS, Shubhorup Biswas and Francis Rhys Ward
11 Feb 2026 21:25 UTC
26 points
0 comments6 min readLW link

Effi­ciently De­tect­ing Hid­den Rea­son­ing with a Small Pre­dic­tor Model

RohanS, Vishnu Vardhan Sai Lanka, yaumeng and daria
13 Jul 2025 16:04 UTC
34 points
3 comments16 min readLW link

Per­spec­tives on Con­tinual Learn­ing: Sur­vey Re­sults and Forecasts

Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward and Seth Herd
24 Jun 2026 16:30 UTC
33 points
0 comments12 min readLW link

Held-out Mon­i­tors Some­times De­grade, Even When Not Trained Against

Joey Yudelson29 Jul 2026 14:56 UTC
62 points
0 comments11 min readLW link

Should We Train Against (CoT) Mon­i­tors?

RohanS23 Apr 2026 19:19 UTC
52 points
7 comments33 min readLW link

Model Or­ganisms of Sand­bag­ging in the Wild

Vladimir Ivanov6 Aug 2026 17:25 UTC
30 points
0 comments8 min readLW link

An­gles of at­tack for con­tinual learn­ing safety

Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward and Seth Herd
16 Jun 2026 16:15 UTC
47 points
0 comments13 min readLW link

How we spent our first two weeks as an in­de­pen­dent AI safety re­search group

RohanS, Rauno Arike and Shubhorup Biswas
11 Aug 2025 19:32 UTC
34 points
0 comments10 min readLW link

We Should Study the Anal­ogy Between Inoc­u­la­tion Prompt­ing Non-Ro­bust­ness, Ne­ga­tion Ne­glect, and Back­door Non-Robustness

Vladimir Ivanov28 May 2026 19:17 UTC
24 points
4 comments4 min readLW link

Aether is hiring tech­ni­cal AI safety researchers

Rauno Arike, RohanS and Shubhorup Biswas
5 Jan 2026 22:27 UTC
22 points
0 comments2 min readLW link
No comments.