INDEX

Explanations

words related to reaching or achieving high positions or quantities

oai_token-act-pair · gpt-3.5-turbo

instances related to maximum values or achievements

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

 Disc

-0.65

 abort

-0.65

 Fisher

-0.64

ibly

-0.64

 actions

-0.61

 contacts

-0.61

 interaction

-0.61

im

-0.60

 interference

-0.60

XY

-0.60

POSITIVE LOGITS

 topped

3.60

 topping

2.21

 surpassed

1.69

 tops

1.57

 dipped

1.50

 garn

1.41

 capped

1.36

 peaked

1.36

 adorned

1.28

 edged

1.28

Activations Density 0.017%

words related to reaching or achieving high positions or quantities

instances related to maximum values or achievements

No Comments

No Known Activations