INDEX

Explanations

information related to television viewership

oai_token-act-pair · gpt-3.5-turbo

mentions of audience viewership and ratings

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 reorgan

-0.71

 Surgery

-0.65

 contingency

-0.65

erald

-0.64

redo

-0.64

orie

-0.62

 Roses

-0.62

 Whites

-0.61

 taxing

-0.61

 Shotgun

-0.59

POSITIVE LOGITS

 viewers

1.01

 viewership

0.97

 tuned

0.92

hip

0.84

wat

0.82

hao

0.79

 tuning

0.77

 viewer

0.75

 listeners

0.75

 "$:/

0.75

Activations Density 0.014%

information related to television viewership

mentions of audience viewership and ratings

No Comments

No Known Activations