INDEX

Explanations

words related to peaks and troughs, particularly focusing on the concept of "troughs"

oai_token-act-pair · gpt-3.5-turbo

references to fluctuations or variations in systems

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

apest

-0.63

ilities

-0.60

uties

-0.55

 reason

-0.53

usual

-0.53

venge

-0.53

 excuse

-0.52

izer

-0.52

 Salam

-0.52

 Wait

-0.51

POSITIVE LOGITS

 through

3.04

 thru

2.48

through

2.36

 Through

2.19

Through

1.93

via

1.75

 trough

1.59

THR

1.30

 throughout

1.09

via

1.05

Activations Density 0.066%

words related to peaks and troughs, particularly focusing on the concept of "troughs"

references to fluctuations or variations in systems

No Comments

No Known Activations