Johannes Treutlein

Karma: 1,534

All opinions are my own. Homepage: johannestreutlein.com

Modifying LLM Beliefs with Synthetic Document Finetuning

RowanWang, Johannes Treutlein, Avery, Ethan Perez, Fabien Roger and Sam Marks

Apr 24, 2025, 9:15 PM

70 points

12 comments2 min readLW link

(alignment.anthropic.com)

Auditing language models for hidden objectives

Sam Marks, Johannes Treutlein, dmz, Sam Bowman, Hoagy, Carson Denison, Kei, 7vik, Akbir Khan, Austin Meek, Euan Ong, Christopher Olah, Fabien Roger, jeanne_, Meg, Drake Thomas, Adam Jermyn, Monte M and evhub

Mar 13, 2025, 7:18 PM

141 points

15 comments13 min readLW link

Alignment Faking in Large Language Models

ryan_greenblatt, evhub, Carson Denison, Benjamin Wright, Fabien Roger, Monte M, Sam Marks, Johannes Treutlein, Sam Bowman and Buck

Dec 18, 2024, 5:19 PM

485 points

75 comments10 min readLW link

Connecting the Dots: LLMs can Infer & Verbalize Latent Structure from Training Data

Johannes Treutlein and Owain_Evans

Jun 21, 2024, 3:54 PM

163 points

13 comments8 min readLW link

(arxiv.org)

Report on modeling evidential cooperation in large worlds

Johannes TreutleinJul 12, 2023, 4:37 PM

45 points

3 comments1 min readLW link

(arxiv.org)

Conditional Prediction with Zero-Sum Training Solves Self-Fulfilling Prophecies

Rubi J. Hudson and Johannes Treutlein

May 26, 2023, 5:44 PM

88 points

13 comments24 min readLW link

Conditioning Predictive Models: Open problems, Conclusion, and Appendix

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 10, 2023, 7:21 PM

36 points

3 comments11 min readLW link

Conditioning Predictive Models: Deployment strategy

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 9, 2023, 8:59 PM

28 points

0 comments10 min readLW link

Conditioning Predictive Models: Interactions with other approaches

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 8, 2023, 6:19 PM

32 points

2 comments11 min readLW link

Conditioning Predictive Models: Making inner alignment as easy as possible

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 7, 2023, 8:04 PM

27 points

2 comments19 min readLW link

Conditioning Predictive Models: The case for competitiveness

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 6, 2023, 8:08 PM

20 points

3 comments11 min readLW link

Conditioning Predictive Models: Outer alignment via careful conditioning

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 2, 2023, 8:28 PM

72 points

15 comments57 min readLW link

Conditioning Predictive Models: Large language models as predictors

evhub, Adam Jermyn, Johannes Treutlein, Rubi J. Hudson and kcwoolverton

Feb 2, 2023, 8:28 PM

88 points

4 comments13 min readLW link

Stop-gradients lead to fixed point predictions

Johannes Treutlein, Caspar Oesterheld, Rubi J. Hudson and Emery Cooper

Jan 28, 2023, 10:47 PM

37 points

2 comments24 min readLW link

Underspecification of Oracle AI

Rubi J. Hudson, Adam Jermyn and Johannes Treutlein

Jan 15, 2023, 8:10 PM

30 points

12 comments19 min readLW link

Proper scoring rules don’t guarantee predicting fixed points

Johannes Treutlein, Rubi J. Hudson and Caspar Oesterheld

Dec 16, 2022, 6:22 PM

79 points

8 comments21 min readLW link

Response to Katja Grace’s AI x-risk counterarguments

Erik Jenner and Johannes Treutlein

19 Oct 2022 1:17 UTC

77 points

18 comments15 min readLW link

Training goals for large language models

Johannes Treutlein18 Jul 2022 7:09 UTC

28 points

5 comments19 min readLW link

Request for input on multiverse-wide superrationality (MSR)

Johannes Treutlein14 Aug 2018 17:29 UTC

18 points

3 comments1 min readLW link

(effective-altruism.com)

A behaviorist approach to building phenomenological bridges

Johannes Treutlein20 Nov 2017 19:36 UTC

4 points

0 comments1 min readLW link

(casparoesterheld.com)