INDEX

Explanations

politically charged terms, such as political affiliations and positions

oai_token-act-pair · gpt-3.5-turbo

references to political affiliations and policies

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĢ

-0.82

\"

-0.76

****************

-0.72

ãĥİ

-0.72

 tonight

-0.72

ULTS

-0.71

""

-0.69

aturday

-0.69

âĹ

-0.69

Picture

-0.66

POSITIVE LOGITS

 themselves

0.93

 herself

0.85

 himself

0.82

 itself

0.77

 elsewhere

0.71

 apiece

0.70

 overseen

0.69

 outright

0.68

.)

0.68

 favoured

0.67

Activations Density 0.768%

politically charged terms, such as political affiliations and positions

references to political affiliations and policies

No Comments

No Known Activations