INDEX

Explanations

locations based on names of places or institutions

oai_token-act-pair · gpt-3.5-turbo

locations followed by the word "based" indicating an organization or group

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Dickinson

-0.82

 Norris

-0.76

 Frost

-0.72

 shade

-0.66

 Farrell

-0.66

 Finn

-0.65

 Levi

-0.65

 Indigo

-0.65

etooth

-0.63

 Grimes

-0.63

POSITIVE LOGITS

based

1.40

themed

1.14

derived

1.12

style

1.10

centric

1.10

oriented

1.09

driven

1.07

sized

1.06

focused

1.05

backed

1.03

Activations Density 0.232%

locations based on names of places or institutions

locations followed by the word "based" indicating an organization or group

No Comments

No Known Activations