INDEX

Explanations

the word "avoided" with varying degrees of emphasis

oai_token-act-pair · gpt-3.5-turbo

instances of the word "avoided" and related terms

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

iop

-1.02

ingham

-0.99

alyst

-0.76

ocrats

-0.71

odore

-0.68

eds

-0.68

 halls

-0.67

opter

-0.67

iate

-0.66

essee

-0.66

POSITIVE LOGITS

 avoided

1.09

 avoids

0.98

avoid

0.97

 avoiding

0.90

 avoidance

0.87

 aver

0.84

 avoid

0.81

 Avoid

0.80

vana

0.78

icho

0.77

Activations Density 0.009%

the word "avoided" with varying degrees of emphasis

instances of the word "avoided" and related terms

No Comments

No Known Activations