INDEX

Explanations

words related to the name "Pandit"

oai_token-act-pair · gpt-3.5-turbo

references to the word "Panda" and its variations, indicating a focus on that topic

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

yr

-0.73

 Jarvis

-0.69

 landfill

-0.68

 relat

-0.67

 brake

-0.64

 Jefferson

-0.64

 divorced

-0.63

 improvement

-0.63

 appliance

-0.62

 steroid

-0.62

POSITIVE LOGITS

 Pand

4.06

Pand

2.33

 Panda

1.62

 Sind

1.12

Zot

1.10

 Gupta

1.06

Rav

1.00

 Pandora

1.00

 Mandal

0.99

Lal

0.99

Activations Density 0.024%

words related to the name "Pandit"

references to the word "Panda" and its variations, indicating a focus on that topic

No Comments

No Known Activations