INDEX

Explanations

words related to asking questions

oai_token-act-pair · gpt-3.5-turbo

instances of the negative contraction "n't"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Scrib

-0.62

 princ

-0.61

Ĥ¬

-0.60

 Gutenberg

-0.59

senal

-0.59

EStreamFrame

-0.57

 halves

-0.57

 elev

-0.56

 protected

-0.56

SAS

-0.56

POSITIVE LOGITS

't

1.34

aught

1.04

aturally

0.94

ÃŃ

0.94

igh

0.93

athan

0.92

ought

0.87

ighed

0.86

augh

0.82

ether

0.82

Activations Density 0.027%

words related to asking questions

instances of the negative contraction "n't"

No Comments

No Known Activations