INDEX

Explanations

mentions of something being highlighted or showcased, possibly in a visual or prominent way

oai_token-act-pair · gpt-3.5-turbo

instances of the word "feature" and its variations in relation to various subjects

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

sis

-1.03

alty

-0.93

hood

-0.78

externalActionCode

-0.77

bell

-0.76

si

-0.70

å¥³

-0.69

iyah

-0.68

trust

-0.67

nai

-0.66

POSITIVE LOGITS

 prominently

1.45

 visuals

0.77

 elements

0.74

 clips

0.72

istors

0.70

leck

0.70

 celebrities

0.70

 snippets

0.69

 rappers

0.69

 artwork

0.68

Activations Density 0.048%

mentions of something being highlighted or showcased, possibly in a visual or prominent way

instances of the word "feature" and its variations in relation to various subjects

No Comments

No Known Activations