INDEX

Explanations

formatting commands or bullet points

oai_token-act-pair · gpt-3.5-turbo

numbered lists or items referencing procedures or steps

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

buck

-0.69

sta

-0.64

 grocer

-0.64

rentices

-0.62

 nurs

-0.62

bern

-0.62

oga

-0.61

 rigging

-0.61

nam

-0.60

azo

-0.59

POSITIVE LOGITS

[+

0.81

————————————————

0.72

eous

0.69

————————

0.69

ãĥīãĥ©ãĤ´ãĥ³

0.68

>]

0.67

 WARRANT

0.67

 Turns

0.65

ãĤ´ãĥ³

0.63

 Prot

0.63

Activations Density 0.027%

formatting commands or bullet points

numbered lists or items referencing procedures or steps

No Comments

No Known Activations