INDEX

Explanations

words related to providing instructions or explanations

oai_token-act-pair · gpt-3.5-turbo

instructions or questions about processes and methods

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

 tert

-0.65

nect

-0.62

entry

-0.61

ãĥĥãĥī

-0.59

 visitation

-0.59

 Bened

-0.58

 Dominican

-0.58

 COVER

-0.57

 consolidation

-0.57

article

-0.57

POSITIVE LOGITS

inki

0.80

called

0.78

irlf

0.71

oward

0.70

ever

0.64

ahs

0.63

ught

0.62

igans

0.62

incible

0.61

fficient

0.60

Activations Density 0.107%

words related to providing instructions or explanations

instructions or questions about processes and methods

No Comments

No Known Activations