INDEX

Explanations

words related to famous individuals or entities

oai_token-act-pair · gpt-3.5-turbo

words or suffixes related to forms or formalities, particularly those that suggest official titles or roles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Decker

-0.69

 Solitaire

-0.65

 Maya

-0.63

âĶĢâĶĢâĶĢâĶĢ

-0.62

FTA

-0.62

 Rico

-0.61

print

-0.60

prints

-0.59

Cu

-0.59

 Marcos

-0.59

POSITIVE LOGITS

orph

1.13

orm

1.07

ont

1.07

orthy

1.04

ophon

0.96

orrow

0.93

endez

0.93

ajor

0.92

etheus

0.92

yss

0.90

Activations Density 0.006%

words related to famous individuals or entities

words or suffixes related to forms or formalities, particularly those that suggest official titles or roles

No Comments

No Known Activations