INDEX

Explanations

the letter 't' appearing with high activations

oai_token-act-pair · gpt-3.5-turbo

instances of the negation "didn't."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 Reduced

-0.64

 Compar

-0.61

itiz

-0.61

 arsen

-0.60

lining

-0.60

edIn

-0.59

Pop

-0.59

soType

-0.59

 Strongh

-0.58

 Dise

-0.58

POSITIVE LOGITS

 bother

0.99

 hesitate

0.91

 necessarily

0.89

 hesitated

0.77

 exactly

0.75

apest

0.75

hes

0.75

actic

0.75

 dare

0.74

ificate

0.74

Activations Density 0.059%

the letter 't' appearing with high activations

instances of the negation "didn't."

No Comments

No Known Activations