INDEX

Explanations

mentions of the word "Hum" with comparably high activations

oai_token-act-pair · gpt-3.5-turbo

references to humor or comedic elements

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

cape

-0.71

 Peaks

-0.71

 ---------

-0.71

 EntityItem

-0.70

ãĤ´ãĥ³

-0.68

å§

-0.65

hips

-0.65

 flagged

-0.64

 Sands

-0.62

 approve

-0.62

POSITIVE LOGITS

pty

1.06

ility

1.04

orously

1.03

iday

1.02

undai

1.01

ankind

0.99

Hum

0.92

Hum

0.88

mers

0.87

obiles

0.86

Activations Density 0.012%

mentions of the word "Hum" with comparably high activations

references to humor or comedic elements

No Comments

No Known Activations