INDEX

Explanations

number sequences with titles such as dates and phrases related to storytelling

oai_token-act-pair · gpt-3.5-turbo

demonstrative pronouns such as "this" and "that"

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

letes

-0.74

enegger

-0.73

ocate

-0.69

lement

-0.67

RIC

-0.67

upon

-0.66

encers

-0.66

gerald

-0.64

avorite

-0.64

blers

-0.62

POSITIVE LOGITS

 occasions

1.39

 occasion

1.25

 behalf

1.17

 basis

1.13

eve

0.94

ilts

0.90

 pretext

0.90

 occas

0.87

 fronts

0.86

demand

0.85

Activations Density 0.231%

number sequences with titles such as dates and phrases related to storytelling

demonstrative pronouns such as "this" and "that"

No Comments

No Known Activations