INDEX

Explanations

references to specific physical spaces or rooms

oai_token-act-pair · gpt-3.5-turbo

references to various types of rooms

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

vous

-0.62

Uz

-0.61

FSA

-0.60

Iss

-0.59

SB

-0.59

Published

-0.59

ÂłÂłÂłÂłÂłÂłÂłÂł

-0.59

 relentless

-0.57

ocrat

-0.56

++++++++++++++++

-0.55

POSITIVE LOGITS

 rooms

1.31

rooms

1.21

 Rooms

1.20

pots

1.10

room

1.00

wagen

0.98

itory

0.96

chool

0.94

hops

0.93

creen

0.89

Activations Density 0.013%

references to specific physical spaces or rooms

references to various types of rooms

No Comments

No Known Activations