INDEX

Explanations

proper names and terms with the root "recogn" followed by certain patterns like "ising" or "isable"

oai_token-act-pair · gpt-3.5-turbo

terms related to recognition and recognizability

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

 Grind

-0.77

 Rail

-0.69

 Tempest

-0.64

 Travel

-0.60

 Diver

-0.59

 ride

-0.59

 Heaven

-0.59

 delaying

-0.59

 hull

-0.59

 Townsend

-0.58

POSITIVE LOGITS

isable

1.22

recogn

1.21

isance

1.21

izable

1.08

ibly

1.07

 recogn

0.99

ably

0.97

ledged

0.96

enance

0.94

orial

0.93

Activations Density 0.008%

proper names and terms with the root "recogn" followed by certain patterns like "ising" or "isable"

terms related to recognition and recognizability

No Comments

No Known Activations