INDEX

Explanations

phrases or words associated with biased reporting

oai_token-act-pair · gpt-3.5-turbo

references to bias in media reporting

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ft

-0.85

clamation

-0.82

ptoms

-0.82

âĢ¢âĢ¢âĢ¢âĢ¢

-0.77

phrase

-0.76

hner

-0.72

chal

-0.72

thur

-0.71

forts

-0.71

mia

-0.70

POSITIVE LOGITS

 biased

1.28

 unbiased

1.08

 impartial

0.97

 biases

0.79

 undermin

0.78

 opinions

0.78

 bias

0.77

 observers

0.77

 viewpoints

0.75

 citiz

0.74

Activations Density 0.014%

phrases or words associated with biased reporting

references to bias in media reporting

No Comments

No Known Activations