INDEX

Explanations

locations or people that are visually captured in images or photographs

oai_token-act-pair · gpt-3.5-turbo

instances of the word "pictured" and its variations

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

 Immunity

-0.63

 attrition

-0.62

 increment

-0.61

 margin

-0.61

 acquisition

-0.57

Â§

-0.57

escal

-0.57

 adversity

-0.55

termination

-0.55

 negotiation

-0.55

POSITIVE LOGITS

 pictured

3.92

pictured

2.14

 photographed

1.99

 Pict

1.67

 depicted

1.62

Pict

1.38

 shown

1.32

shown

1.30

 described

1.19

 circled

1.19

Activations Density 0.015%

locations or people that are visually captured in images or photographs

instances of the word "pictured" and its variations

No Comments

No Known Activations