INDEX

Explanations

words related to speculation, analysis, and trends

oai_token-act-pair · gpt-3.5-turbo

language indicating opinions or predictions

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

?",

-0.60

)",

-0.59

"},

-0.57

 Femin

-0.57

 reply

-0.53

 poems

-0.53

Nig

-0.53

,"

-0.53

 poem

-0.52

",

-0.52

POSITIVE LOGITS

 expect

0.76

etheless

0.71

 hoping

0.66

 anticipate

0.62

 nonetheless

0.61

ivably

0.60

fy

0.59

 caution

0.58

 anecd

0.58

 unlikely

0.58

Activations Density 0.965%

words related to speculation, analysis, and trends

language indicating opinions or predictions

No Comments

No Known Activations