INDEX

Explanations

political titles and names paired with statements

oai_token-act-pair · gpt-3.5-turbo

occurrences of references to senators in the text

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

 traffickers

-0.74

 scoop

-0.74

zsche

-0.71

 underpin

-0.68

 cradle

-0.67

hiba

-0.66

 varied

-0.65

 meanings

-0.63

obyl

-0.63

 mosqu

-0.62

POSITIVE LOGITS

 Olympia

0.98

 Lamar

0.85

 Majority

0.84

 Barack

0.84

PAC

0.84

Len

0.83

 Louie

0.82

 Dianne

0.82

Yon

0.82

 Jesse

0.81

Activations Density 0.038%

political titles and names paired with statements

occurrences of references to senators in the text

No Comments

No Known Activations