INDEX

Explanations

the word "tre" with high activation values

oai_token-act-pair · gpt-3.5-turbo

references to trees

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Madden

-0.76

 Robots

-0.72

 Mech

-0.68

 Jacobs

-0.68

icka

-0.66

 Scarborough

-0.66

obin

-0.65

 Rivera

-0.64

 Answer

-0.64

 Abstract

-0.63

POSITIVE LOGITS

tre

4.09

tre

2.25

Tre

2.04

Tre

1.85

 trem

1.47

TRE

1.46

 quadru

1.26

 Trem

1.08

tul

1.06

tra

1.06

Activations Density 0.022%

the word "tre" with high activation values

references to trees

No Comments

No Known Activations