INDEX

Explanations

NB: The given text for Neuron 4 is very long to interpret accurately by manual analysis. You might need to consider a more specific subset of the text to precisely determine what the neuron is looking for. Feel free to provide a subset that you'd like help with!

oai_token-act-pair · gpt-3.5-turbo

phrases related to legal or evidentiary assessments

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

ãĤ©

-0.55

ãĥ´

-0.55

ãĤ£

-0.48

Lots

-0.47

 Courtesy

-0.47

 earthqu

-0.46

âĶ

-0.44

 beware

-0.44

 Sounds

-0.43

LOT

-0.43

POSITIVE LOGITS

nor

1.74

 anymore

1.42

 whatsoever

1.37

nor

1.22

or

1.19

 anywhere

1.08

 anything

0.98

 anytime

0.97

 either

0.97

any

0.94

Activations Density 1.193%

NB: The given text for Neuron 4 is very long to interpret accurately by manual analysis. You might need to consider a more specific subset of the text to precisely determine what the neuron is looking for. Feel free to provide a subset that you'd like help with!

phrases related to legal or evidentiary assessments

No Comments

No Known Activations

NB: The given text for Neuron 4 is very long to interpret accurately by manual analysis. You might need to consider a more specific subset of the text to precisely determine what the neuron is looking for. Feel free to provide a subset that you'd like help with!

phrases related to legal or evidentiary assessments

No Comments

No Known Activations