INDEX

Explanations

words related to architectural features, particularly lobbies

oai_token-act-pair · gpt-3.5-turbo

references to lobbies and lobbying activity

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

âĵĺ

-0.67

DIT

-0.65

videos

-0.65

vous

-0.64

 Constantin

-0.64

tics

-0.64

ebook

-0.64

fal

-0.63

 Crus

-0.63

 Reincarn

-0.63

POSITIVE LOGITS

 lobb

1.07

 lobby

1.00

 Lobby

0.90

ature

0.88

creen

0.87

naires

0.85

heit

0.83

ounge

0.83

busters

0.79

uese

0.77

Activations Density 0.010%

words related to architectural features, particularly lobbies

references to lobbies and lobbying activity

No Comments

No Known Activations