Subliminal learning: Models transmit behaviors via hidden signals in data (alignment.anthropic.com) | HN Screenshots

YHacker Newsscreenshots

Screenshot of alignment.anthropic.com: Subliminal learning: Models transmit behaviors via hidden signals in data

First screen, captured 1 year ago.

Looks similar

Screenshot of machinelearning.apple.com: FastVLM: Efficient Vision Encoding for Vision Language Models

FastVLM: Efficient Vision Encoding for Vision Language Models (machinelearning.apple.com)

82 points by 2bit 1 year ago | 5 comments

Screenshot of lightly.ai: Segment Anything Model and Friends

Segment Anything Model and Friends (lightly.ai)

95 points by sauravmaheshkar 2 years ago | 14 comments

Screenshot of assemblyai.com: How RLHF Preference Model Tuning Works (and How Things May Go Wrong)

How RLHF Preference Model Tuning Works (and How Things May Go Wrong)(assemblyai.com)

88 points by dylanbfox 3 years ago | 4 comments

Screenshot of sebastianraschka.com: Optimizing LLMs from a Dataset Perspective

Optimizing LLMs from a Dataset Perspective (sebastianraschka.com)

126 points by alexmolas 3 years ago | 23 comments

Screenshot of joecarlsmith.com: Takes on "Alignment Faking in Large Language Models"

Takes on "Alignment Faking in Large Language Models"(joecarlsmith.com)

6 points by surprisetalk 1 year ago | 1 comments

Screenshot of quantamagazine.org: Parallels in the ways that humans and ML models acquire language skills

Parallels in the ways that humans and ML models acquire language skills (quantamagazine.org)

94 points by theafh 3 years ago | 24 comments