INDEX

Explanations

keywords related to identification or recognizing individuals or entities

oai_token-act-pair · gpt-3.5-turbo

mentions of identification or identifiers related to specific entities or concepts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 drip

-0.79

 breeze

-0.76

 bribery

-0.71

zzle

-0.71

 bribe

-0.70

 puff

-0.68

 fallout

-0.67

 Fortune

-0.67

arella

-0.67

akura

-0.67

POSITIVE LOGITS

 Ident

3.54

Ident

3.22

 identifiers

1.76

 ident

1.71

 Catal

1.69

 Identification

1.58

 identifier

1.56

Catal

1.28

ident

1.22

 Identified

1.19

Activations Density 0.055%

keywords related to identification or recognizing individuals or entities

mentions of identification or identifiers related to specific entities or concepts

No Comments

No Known Activations