INDEX

Explanations

phrases related to persuading or convincing someone to take a specific action

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate an effort to persuade or influence someone

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

advertising

-0.70

é¾įå¥ĳå£«

-0.66

 Judging

-0.66

 Effects

-0.64

 contrasted

-0.63

Compar

-0.63

 inferred

-0.62

 Reports

-0.61

 Chall

-0.61

runtime

-0.60

POSITIVE LOGITS

 cooperate

1.09

 accept

0.99

 obey

0.99

 behave

0.97

 agree

0.96

 commit

0.96

 comply

0.95

 participate

0.94

 soften

0.93

 admit

0.93

Activations Density 0.107%

phrases related to persuading or convincing someone to take a specific action

phrases that indicate an effort to persuade or influence someone

No Comments

No Known Activations