INDEX

Explanations

action-related words related to physical combat, such as "blows" and "punches"

oai_token-act-pair · gpt-3.5-turbo

terms related to physical confrontations or violence

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

 Slate

-0.72

alph

-0.65

ramid

-0.64

 overlay

-0.63

 Priority

-0.63

 track

-0.61

ument

-0.61

skip

-0.61

 loop

-0.60

pip

-0.59

POSITIVE LOGITS

 blows

3.28

 punches

2.33

 throws

1.87

 knocks

1.51

 kicks

1.51

 insults

1.43

 glances

1.41

 strikes

1.36

 embraces

1.33

 kisses

1.32

Activations Density 0.023%

action-related words related to physical combat, such as "blows" and "punches"

terms related to physical confrontations or violence

No Comments

No Known Activations