INDEX

Explanations

the name "Be"

oai_token-act-pair · gpt-3.5-turbo

references to the word "Be" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 firsthand

-0.73

 totality

-0.73

intendent

-0.69

 resonance

-0.69

 flare

-0.67

 partName

-0.67

 glide

-0.67

 lapse

-0.65

ilver

-0.64

 pole

-0.64

POSITIVE LOGITS

yond

1.17

hemoth

1.17

aum

1.15

arers

1.12

acons

1.05

ech

1.04

ards

1.04

cker

1.02

friends

1.02

arer

1.01

Activations Density 0.024%

the name "Be"

references to the word "Be" in various contexts

No Comments

No Known Activations