INDEX

Explanations

proper nouns and terms, possibly related to news articles or publications

oai_token-act-pair · gpt-3.5-turbo

references to value or importance, particularly in the context of art and collections

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

unia

-0.62

keeper

-0.60

Ö¼

-0.58

 recons

-0.57

issa

-0.55

gue

-0.55

isse

-0.55

afety

-0.54

ienne

-0.54

arden

-0.54

POSITIVE LOGITS

 respectively

0.81

 apiece

0.77

 nominees

0.77

 Lists

0.77

 nominations

0.77

 PACs

0.76

 list

0.75

ategor

0.73

 trio

0.72

 eclectic

0.71

Activations Density 1.188%

proper nouns and terms, possibly related to news articles or publications

references to value or importance, particularly in the context of art and collections

No Comments

No Known Activations