INDEX

Explanations

names of characters

oai_token-act-pair · gpt-3.5-turbo

the word "named" followed by various character names

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

EMS

-0.86

keyes

-0.74

romy

-0.74

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

-0.73

qua

-0.72

irth

-0.71

PLIED

-0.68

sth

-0.66

zers

-0.65

acia

-0.65

POSITIVE LOGITS

 named

0.79

 surn

0.78

 aliases

0.78

named

0.77

paces

0.72

 names

0.72

milo

0.69

calling

0.68

ames

0.67

 Alvin

0.67

Activations Density 0.015%

names of characters

the word "named" followed by various character names

No Comments

No Known Activations