INDEX

Explanations

mentions of specific durations of time, particularly "hour."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "hour" used in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

LECT

-0.78

imaru

-0.70

ortium

-0.69

raviolet

-0.65

liga

-0.65

reditary

-0.65

Shock

-0.64

 Causes

-0.62

ce

-0.61

 Bind

-0.61

POSITIVE LOGITS

glass

1.24

endment

0.94

 ILCS

0.89

bnb

0.77

ï¸

0.77

Ago

0.75

neapolis

0.72

 mins

0.72

worm

0.71

istance

0.70

Activations Density 0.017%

mentions of specific durations of time, particularly "hour."

instances of the word "hour" used in various contexts

No Comments

No Known Activations