INDEX

Explanations

the word "ington" at the end of place names

oai_token-act-pair · gpt-3.5-turbo

mentions of locations in the form of place names

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 diver

-0.60

 point

-0.59

 general

-0.58

 contam

-0.58

 adherence

-0.58

 GENERAL

-0.57

 fusion

-0.57

 barrier

-0.56

 transports

-0.55

 decl

-0.55

POSITIVE LOGITS

ington

4.87

INGTON

2.15

ingham

1.32

stown

1.29

lington

1.17

hester

1.17

mington

1.15

erton

1.14

inton

1.09

erness

1.09

Activations Density 0.009%

the word "ington" at the end of place names

mentions of locations in the form of place names

No Comments

No Known Activations