INDEX

Explanations

words that are not relevant in conveying a specific context or meaning but appear repeatedly

oai_token-act-pair · gpt-3.5-turbo

phrases related to email communications and actions

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 painting

-0.68

 Tomb

-0.68

aterasu

-0.67

 inciner

-0.64

cut

-0.63

 stomp

-0.63

 residency

-0.61

 Flam

-0.61

 haircut

-0.60

 Shrine

-0.60

POSITIVE LOGITS

Disclaimer

1.35

Contribut

1.07

Auth

1.06

Feel

1.05

Subscribe

1.04

Comments

1.04

Author

1.04

Copyright

1.00

Links

0.97

Thank

0.96

Activations Density 0.443%

words that are not relevant in conveying a specific context or meaning but appear repeatedly

phrases related to email communications and actions

No Comments

No Known Activations