INDEX

Explanations

words related to political, cultural, and historical terms specific to certain regions

oai_token-act-pair · gpt-3.5-turbo

references to a specific entity or concept, particularly one that is repeated frequently and relevant in the context

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 depress

-0.64

 Philippines

-0.64

 lead

-0.63

 bacon

-0.61

 kill

-0.60

 Firefox

-0.58

 sensitive

-0.57

 flight

-0.57

 Johns

-0.57

 Chester

-0.56

POSITIVE LOGITS

ti

4.53

tu

1.77

tis

1.73

ta

1.60

Ti

1.58

TI

1.45

si

1.39

tin

1.34

shi

1.32

1.26

Activations Density 0.011%

words related to political, cultural, and historical terms specific to certain regions

references to a specific entity or concept, particularly one that is repeated frequently and relevant in the context

No Comments

No Known Activations