INDEX

Explanations

words related to specific cultural terms and names

oai_token-act-pair · gpt-3.5-turbo

references to specific numerical values or coded identifiers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

HI

-0.70

 unden

-0.64

 pasture

-0.63

 Wonderland

-0.63

 Holden

-0.62

 Dele

-0.61

Race

-0.60

 mouse

-0.59

ItemThumbnailImage

-0.58

 psychiat

-0.58

POSITIVE LOGITS

etus

0.82

uca

0.72

unic

0.71

Äĩ

0.71

ilver

0.68

jen

0.66

Lenin

0.65

otech

0.65

ensis

0.65

vu

0.63

Activations Density 0.537%

words related to specific cultural terms and names

references to specific numerical values or coded identifiers

No Comments

No Known Activations