INDEX

Explanations

proper nouns or names that have been given a specific nickname or title

oai_token-act-pair · gpt-3.5-turbo

the word "dubbed" related to various subjects or entities

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

ramid

-0.75

--+

-0.67

adra

-0.61

 Restaur

-0.61

itars

-0.61

ctrl

-0.60

=-=-

-0.60

rogens

-0.60

ija

-0.60

eties

-0.59

POSITIVE LOGITS

 dubbed

0.81

selves

0.74

iously

0.72

ãĥĩ

0.71

phas

0.70

é¾

0.68

"@

0.67

"#

0.67

"<

0.65

icut

0.64

Activations Density 0.024%

proper nouns or names that have been given a specific nickname or title

the word "dubbed" related to various subjects or entities

No Comments

No Known Activations