INDEX

Explanations

phrases or terms relating to official announcements or statements

oai_token-act-pair · gpt-3.5-turbo

sequences of capital letters, likely indicative of abbreviations or acronyms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ÄŁ

-0.63

cause

-0.61

 dstg

-0.60

awatts

-0.59

 Liberties

-0.57

bows

-0.56

 pretext

-0.56

bridge

-0.56

roc

-0.56

ks

-0.55

POSITIVE LOGITS

INS

1.21

ISH

1.18

UTE

1.13

EMENT

1.12

ERY

1.12

ITION

1.12

ING

1.11

ULL

1.11

ILL

1.10

ERS

1.09

Activations Density 0.096%

phrases or terms relating to official announcements or statements

sequences of capital letters, likely indicative of abbreviations or acronyms

No Comments

No Known Activations