bearseascape

2025-06-21 12:18

Commented: "AbsenceBench: Language models can't tell what's missing"

The madlib like sentences approach is actually how masked token prediction works! It was one of the pretraining tasks for BERT, but nowadays I think all (?) LLMs are trained with next token prediction instead.

2025-04-14 6:22

Submitted: "MooseAgent: A LLM Based Multi-Agent Framework for Automating Moose Simulation"

13 points0 commentsarxiv.org

The Finite Element Method (FEM) is widely used in engineering and scientific computing, but its pre-processing, solver configuration, and post-processing stages are often time-consuming and require…

2025-03-27 3:06

Submitted: "Automated Researchers Can Subtly Sandbag"

2 points0 commentsalignment.anthropic.com

tl;dr When prompted, current models can sandbag ML experiments and research decisions without being detected by zero-shot prompted monitors. Claude 3.5 Sonnet (new) can only sandbag effectively when…

2025-03-27 4:11

Submitted: "Auditing Language Models for Hidden Objectives"

1 points0 commentswww.anthropic.com

A collaboration between Anthropic's Alignment Science and Interpretability teams

2025-03-27 3:58

Submitted: "Policy for LLM Writing on LessWrong"

2 points0 commentswww.lesswrong.com

LessWrong has been receiving an increasing number of posts and comments that look like they might be LLM-written or partially-LLM-written, so we're a…

Hacker News

bearseascape

48

2023-04-05

Recent Activity

Commented: "AbsenceBench: Language models can't tell what's missing"

Submitted: "MooseAgent: A LLM Based Multi-Agent Framework for Automating Moose Simulation"

Submitted: "Automated Researchers Can Subtly Sandbag"

Submitted: "Auditing Language Models for Hidden Objectives"

Submitted: "Policy for LLM Writing on LessWrong"

HackerNews