INDEX

Explanations

keywords related to negative or unfavorable situations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "bad."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ĸļ

-0.76

 Lauder

-0.71

ittees

-0.70

 Pavilion

-0.67

chwitz

-0.67

illon

-0.66

channelAvailability

-0.63

brates

-0.62

racuse

-0.62

olate

-0.62

POSITIVE LOGITS

dies

1.40

gered

1.34

ger

1.23

die

1.17

gers

1.16

asses

1.10

dest

1.06

ged

1.04

 apples

1.02

ges

1.02

Activations Density 0.047%

keywords related to negative or unfavorable situations

instances of the word "bad."

No Comments

No Known Activations