INDEX

Explanations

keywords related to drawing attention or focus

oai_token-act-pair · gpt-3.5-turbo

instances of the word "attention" being mentioned 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

 halves

-0.80

tre

-0.65

 Mehran

-0.65

 Yugoslavia

-0.62

 Recon

-0.62

 Staten

-0.61

 Penal

-0.61

 Townsend

-0.60

imar

-0.60

\\\\\\\\

-0.60

POSITIVE LOGITS

estinal

1.01

 attention

0.93

orial

0.86

ively

0.86

arios

0.84

 Attention

0.84

 scrutiny

0.76

stadt

0.74

ibility

0.74

naire

0.73

Activations Density 0.019%

keywords related to drawing attention or focus

instances of the word "attention" being mentioned in various contexts

No Comments

No Known Activations