INDEX

Explanations

phrases related to political news or events

oai_token-act-pair · gpt-3.5-turbo

brackets and punctuation

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

 gems

-0.67

 fandom

-0.66

ties

-0.63

 assemblies

-0.62

 unaccompanied

-0.61

papers

-0.60

stood

-0.60

 disciplines

-0.58

 joints

-0.58

 setting

-0.58

POSITIVE LOGITS

 WATCHED

1.05

 Continued

1.03

 Expand

0.88

Photos

0.85

ccording

0.85

Read

0.83

Meanwhile

0.79

Related

0.78

Attempts

0.77

 Advertisement

0.77

Activations Density 0.194%

phrases related to political news or events

brackets and punctuation

No Comments

No Known Activations