INDEX

Explanations

the word "wrath"

oai_token-act-pair · gpt-3.5-turbo

references to the abbreviation "WR" 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 @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

croft

-0.82

 Stam

-0.66

anwhile

-0.66

HAEL

-0.66

ãģ¦

-0.65

å§«

-0.64

 Tonight

-0.63

RON

-0.63

DEA

-0.62

ensis

-0.61

POSITIVE LOGITS

angler

1.20

ights

1.06

wr

1.06

inkle

1.05

abbit

0.97

angling

0.96

inker

0.95

angle

0.93

anger

0.91

heed

0.88

Activations Density 0.015%

the word "wrath"

references to the abbreviation "WR" and related terms

No Comments

No Known Activations