INDEX

Explanations

phrases related to news events or incidents

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "as."

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

eeee

-0.78

 Rebell

-0.68

Feel

-0.65

Gender

-0.64

ivas

-0.64

eous

-0.64

enne

-0.63

Reply

-0.63

 COUR

-0.63

leness

-0.63

POSITIVE LOGITS

pired

1.04

phy

0.99

 opposed

0.96

 well

0.94

pects

0.90

part

0.90

piring

0.89

ynchron

0.89

 soon

0.89

pires

0.85

Activations Density 0.213%

phrases related to news events or incidents

occurrences of the word "as."

No Comments

No Known Activations