INDEX

Explanations

sequences of characters or symbols that may not form meaningful words or concepts

oai_token-act-pair · gpt-3.5-turbo

sequences of numerical or coded data

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

onential

-0.54

vernment

-0.53

icipated

-0.52

ANG

-0.50

reau

-0.50

iaries

-0.48

actory

-0.46

 glim

-0.46

ãĤ¤ãĥĪ

-0.46

winner

-0.46

POSITIVE LOGITS

ĸļ

0.51

pell

0.47

cheat

0.46

deen

0.46

 tears

0.45

stress

0.45

stice

0.45

 spat

0.44

Ven

0.44

Ort

0.43

Activations Density 1.458%

sequences of characters or symbols that may not form meaningful words or concepts

sequences of numerical or coded data

No Comments

No Known Activations