INDEX

Explanations

words related to political figures and administrative posts

oai_token-act-pair · gpt-3.5-turbo

names of individuals and locations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Seym

-0.79

matic

-0.78

 Pend

-0.75

 Ambro

-0.74

ruciating

-0.72

matically

-0.71

 Able

-0.70

matical

-0.70

TPPStreamerBot

-0.68

oga

-0.68

POSITIVE LOGITS

enos

1.08

eny

1.02

eno

0.84

arios

0.79

chant

0.76

loe

0.74

cyclopedia

0.74

zyme

0.71

semble

0.69

ymes

0.68

Activations Density 0.014%

words related to political figures and administrative posts

names of individuals and locations

No Comments

No Known Activations