INDEX

Explanations

names of people, potentially associated with news or media

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "Ar"

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

 fold

-0.65

 bell

-0.58

 sting

-0.58

 comma

-0.55

 shorth

-0.55

 panic

-0.54

oway

-0.53

ĵĺ

-0.53

 Staten

-0.53

ĨĴ

-0.53

POSITIVE LOGITS

rival

1.44

duino

1.38

paio

1.35

gent

1.27

thritis

1.21

lington

1.20

ranged

1.20

ctic

1.20

rang

1.15

ithmetic

1.15

Activations Density 0.018%

names of people, potentially associated with news or media

occurrences of the word "Ar"

No Comments

No Known Activations