INDEX

Explanations

phrases related to terrorist attacks

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ãĤ¦ãĤ¹

-0.85

igmat

-0.82

ripp

-0.81

milo

-0.75

00200000

-0.74

uid

-0.73

ById

-0.72

shown

-0.72

catentry

-0.71

ozo

-0.71

POSITIVE LOGITS

 embassies

1.11

 civilians

1.05

 innocent

1.02

 churches

1.01

 mosques

0.98

 synagogue

0.94

 weddings

0.93

 Israelis

0.92

 pedestrians

0.91

 offices

0.91

Activations Density 0.228%

phrases related to terrorist attacks

No Comments

No Known Activations