INDEX

Explanations

words related to locations, potentially in a news or political context

oai_token-act-pair · gpt-3.5-turbo

occurrences of the substring "uz", likely related to the context of a specific location or entity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Siren

-0.79

Redditor

-0.69

 Scandinavian

-0.66

 whistle

-0.65

ÂłÂłÂłÂł

-0.63

Proof

-0.62

åĤ

-0.61

 Silent

-0.61

 Planetary

-0.61

Child

-0.60

POSITIVE LOGITS

uz

1.45

hou

1.03

olini

0.98

achev

0.98

uba

0.94

alez

0.93

hao

0.93

arro

0.93

uh

0.92

oru

0.92

Activations Density 0.005%

words related to locations, potentially in a news or political context

occurrences of the substring "uz", likely related to the context of a specific location or entity

No Comments

No Known Activations