INDEX

Explanations

Seems like neuron 4 is activating for misspelled variations of the word "somewhat"

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "somewhat" and related variations

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

gio

-0.73

 Prosecut

-0.72

holm

-0.71

Kh

-0.71

 Feld

-0.68

Reviewer

-0.67

 Luxembourg

-0.65

xual

-0.65

cv

-0.63

selection

-0.62

POSITIVE LOGITS

ety

1.08

ently

1.02

ency

1.01

hest

0.97

ards

0.97

atell

0.95

actory

0.91

ato

0.91

ences

0.90

eday

0.90

Activations Density 0.028%

Seems like neuron 4 is activating for misspelled variations of the word "somewhat"

occurrences of the word "somewhat" and related variations

No Comments

No Known Activations