INDEX

Explanations

phrases related to teasing or making fun of someone or something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "poke" in various forms

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ingham

-0.95

icter

-0.73

inction

-0.71

recy

-0.71

ccording

-0.67

 Concord

-0.67

ophers

-0.66

funding

-0.65

âĢ¢âĢ¢âĢ¢âĢ¢

-0.62

actions

-0.62

POSITIVE LOGITS

 poking

1.10

 poke

0.96

cheon

0.81

 poked

0.81

ature

0.79

geon

0.77

gery

0.74

balls

0.73

 peek

0.71

 holes

0.71

Activations Density 0.016%

phrases related to teasing or making fun of someone or something

instances of the word "poke" in various forms

No Comments

No Known Activations