INDEX

Explanations

Japanese names, potentially related to locations or individuals

oai_token-act-pair · gpt-3.5-turbo

the names of individuals associated with a specific context or event

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Rudd

-0.66

 Lynd

-0.65

 markup

-0.64

 Redux

-0.63

 residual

-0.62

 pollut

-0.61

 pasture

-0.61

 forwards

-0.59

 straw

-0.58

 view

-0.57

POSITIVE LOGITS

achi

4.67

iba

1.40

hiba

1.28

achu

1.20

uku

1.13

agi

1.08

aeda

1.07

ichi

1.03

anka

1.01

atsu

1.01

Activations Density 0.007%

Japanese names, potentially related to locations or individuals

the names of individuals associated with a specific context or event

No Comments

No Known Activations