INDEX

Explanations

the first instance of a particular event being reported

oai_token-act-pair · gpt-3.5-turbo

instances of the word "first" indicating initial reports or occurrences

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

 Gould

-0.65

ourge

-0.62

sav

-0.61

 Tant

-0.60

 Doomsday

-0.60

 Canaver

-0.58

âĹ¼

-0.58

 Savage

-0.58

holes

-0.57

gery

-0.57

POSITIVE LOGITS

 responders

1.13

 baseman

1.09

 blush

0.81

 glance

0.80

 ancest

0.79

 impressions

0.77

arthed

0.73

 glim

0.73

ebin

0.72

 impression

0.72

Activations Density 0.045%

the first instance of a particular event being reported

instances of the word "first" indicating initial reports or occurrences

No Comments

No Known Activations