INDEX

Explanations

phrases related to news events and accusations

oai_token-act-pair · gpt-3.5-turbo

references to events involving public figures and crises

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

pmwiki

-0.70

 guides

-0.64

Enlarge

-0.63

 fellows

-0.62

 sqor

-0.62

ategory

-0.61

 conductor

-0.60

ãĤ¦ãĤ¹

-0.60

 Amit

-0.59

 rails

-0.59

POSITIVE LOGITS

ardless

0.80

kt

0.75

eries

0.75

eng

0.71

sych

0.68

ickr

0.67

ansion

0.66

ikan

0.66

 flooding

0.66

fw

0.66

Activations Density 0.000%

phrases related to news events and accusations

references to events involving public figures and crises

No Comments

No Known Activations