INDEX

Explanations

words containing the subsequence "mi"

oai_token-act-pair · gpt-3.5-turbo

instances of the token "mi" indicating a focus on multimedia or visual elements

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

 blasphemy

-0.69

 categor

-0.65

 Asgard

-0.65

 frog

-0.62

 crowds

-0.62

 bans

-0.60

 buttocks

-0.60

 buck

-0.59

 Scourge

-0.59

 sting

-0.58

POSITIVE LOGITS

mi

4.52

Mi

1.69

mi

1.64

mia

1.55

mie

1.44

mic

1.42

MI

1.33

Mi

1.28

igi

1.24

imi

1.24

Activations Density 0.015%

words containing the subsequence "mi"

instances of the token "mi" indicating a focus on multimedia or visual elements

No Comments

No Known Activations