INDEX

Explanations

It's challenging to determine the specific theme based on the provided activations for neuron 4. Let's try to filter out the pertinent information which will be essential to summarize

oai_token-act-pair · gpt-3.5-turbo

sentence endings and punctuation

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 feder

-0.72

 adversaries

-0.68

 analysts

-0.67

 insurg

-0.67

 credential

-0.66

 assass

-0.66

 scholars

-0.65

 assassination

-0.65

 contextual

-0.65

 mascul

-0.65

POSITIVE LOGITS

 Ended

1.50

 Turns

1.45

 Needless

1.41

 Afterwards

1.39

 Luckily

1.36

 Anyway

1.32

 Thankfully

1.25

 Then

1.24

 Eventually

1.22

 Upon

1.20

Activations Density 0.229%

It's challenging to determine the specific theme based on the provided activations for neuron 4. Let's try to filter out the pertinent information which will be essential to summarize

sentence endings and punctuation

No Comments

No Known Activations

It's challenging to determine the specific theme based on the provided activations for neuron 4. Let's try to filter out the pertinent information which will be essential to summarize

sentence endings and punctuation

No Comments

No Known Activations