INDEX

Explanations

phrases related to humor or jokes

oai_token-act-pair · gpt-3.5-turbo

references to humor or comedic content

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

ainer

-1.02

eded

-0.82

enfranch

-0.82

eding

-0.78

ignt

-0.77

chwitz

-0.76

ilings

-0.75

ensional

-0.74

apers

-0.74

redit

-0.73

POSITIVE LOGITS

netflix

0.92

 funny

0.82

GIF

0.79

 Laugh

0.75

balls

0.75

ness

0.75

 comedy

0.74

glers

0.72

 banter

0.72

 anecdotes

0.71

Activations Density 0.021%

phrases related to humor or jokes

references to humor or comedic content

No Comments

No Known Activations