INDEX

Explanations

instances of requests for interaction such as replies or deletions in online conversations

oai_token-act-pair · gpt-3.5-turbo

frequent usage of the word "Reply" in a conversational context

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

 severity

-0.71

isks

-0.69

 erupt

-0.66

ometers

-0.66

ometer

-0.66

 weave

-0.63

 bombard

-0.63

 seriousness

-0.62

anship

-0.62

 rounding

-0.61

POSITIVE LOGITS

 Reply

3.93

Reply

1.63

 Delete

1.33

eln

0.91

 References

0.90

alus

0.89

 UNCLASSIFIED

0.83

 Chap

0.80

 Quote

0.79

Frag

0.77

Activations Density 0.013%

instances of requests for interaction such as replies or deletions in online conversations

frequent usage of the word "Reply" in a conversational context

No Comments

No Known Activations