INDEX

Explanations

names of news agencies and possibly associated types of content

oai_token-act-pair · gpt-3.5-turbo

references to the Associated Press

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 ment

-0.79

gaard

-0.66

 stabil

-0.65

pants

-0.64

stood

-0.63

 trout

-0.62

tty

-0.62

abiding

-0.61

 Founders

-0.61

 gluten

-0.60

POSITIVE LOGITS

PLIED

1.09

TN

1.01

PLIC

1.00

PLE

0.96

ocalypse

0.94

OE

0.89

PLA

0.88

CBC

0.87

rison

0.86

HAEL

0.86

Activations Density 0.011%

names of news agencies and possibly associated types of content

references to the Associated Press

No Comments

No Known Activations