INDEX

Explanations

questions

oai_token-act-pair · gpt-3.5-turbo

questions or prompts that invite audience feedback or opinion

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

opes

-0.77

bidden

-0.65

 thieves

-0.64

rupulous

-0.64

pher

-0.64

 fraternity

-0.62

 intermedi

-0.62

 dece

-0.62

 initialization

-0.62

 unsuspecting

-0.61

POSITIVE LOGITS

 Discuss

1.35

 Submit

1.30

 Comment

1.26

 Feel

1.20

 Leave

1.20

 Share

1.20

 Please

1.19

Let

1.18

 Would

1.18

 Email

1.14

Activations Density 0.090%

questions

questions or prompts that invite audience feedback or opinion

No Comments

No Known Activations