INDEX

Explanations

references to physical locations, specifically company headquarters

oai_token-act-pair · gpt-3.5-turbo

mentions of "headquarters."

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

ebook

-0.83

spir

-0.76

flix

-0.75

ef

-0.72

haps

-0.71

erer

-0.71

formed

-0.70

oise

-0.70

nic

-0.69

POSITIVE LOGITS

 headquarters

1.02

 building

0.79

enture

0.78

eteria

0.77

 location

0.77

 Headquarters

0.76

 glim

0.76

 reception

0.75

 courtyard

0.73

 address

0.73

Activations Density 0.027%

references to physical locations, specifically company headquarters

mentions of "headquarters."

No Comments

No Known Activations