INDEX

Explanations

numerical quantities or emphasis

oai_token-act-pair · gpt-3.5-turbo

phrases indicating personal perceptions or interpretations

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

 impl

-0.63

igr

-0.63

worthy

-0.61

esian

-0.61

 endeavor

-0.59

abin

-0.58

worthiness

-0.57

 endeavour

-0.57

ridge

-0.56

 demol

-0.55

POSITIVE LOGITS

rid

1.18

tin

0.94

 bored

0.92

DragonMagazine

0.86

 lucky

0.84

 tired

0.81

TING

0.78

cloneembedreportprint

0.78

 reimb

0.77

 rewarded

0.75

Activations Density 0.120%

numerical quantities or emphasis

phrases indicating personal perceptions or interpretations

No Comments

No Known Activations