INDEX

Explanations

the word "go" with varying activations

oai_token-act-pair · gpt-3.5-turbo

the word "go" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Aram

-0.76

è¦ļéĨĴ

-0.64

IDENT

-0.62

 internally

-0.61

HH

-0.61

dimension

-0.60

 Jude

-0.60

 Topic

-0.58

Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯Â¯

-0.58

 Hour

-0.58

POSITIVE LOGITS

ggle

0.96

lems

0.95

ogl

0.93

ethe

0.91

verning

0.89

go

0.89

ogly

0.87

vernment

0.85

ffic

0.83

ardless

0.81

Activations Density 0.007%

the word "go" with varying activations

the word "go" in various contexts

No Comments

No Known Activations