INDEX

Explanations

It seems that neuron 4 is having trouble identifying a specific pattern in the text provided as there are many different characters and seemingly random activations

oai_token-act-pair · gpt-3.5-turbo

characters from various alphabets and symbols

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

anwhile

-0.99

 msec

-0.82

theless

-0.81

ftime

-0.78

 agre

-0.75

nyder

-0.73

abase

-0.71

 dope

-0.70

espie

-0.70

 cocaine

-0.69

POSITIVE LOGITS

¥

1.69

ı

1.58

Į

1.52

İ

1.51

Ł

1.51

»

1.50

Ī

1.49

²

1.49

´

1.48

º

1.47

Activations Density 0.017%

It seems that neuron 4 is having trouble identifying a specific pattern in the text provided as there are many different characters and seemingly random activations

characters from various alphabets and symbols

No Comments

No Known Activations