INDEX

Explanations

words related to urging or advising actions or behaviors

oai_token-act-pair · gpt-3.5-turbo

references to groups of people being encouraged or advised to take specific actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

atile

-0.67

mys

-0.64

ELD

-0.61

 Built

-0.58

acebook

-0.54

 certs

-0.53

Rated

-0.51

ILA

-0.50

ater

-0.50

anka

-0.50

POSITIVE LOGITS

 beware

1.08

to

1.01

 against

0.92

not

0.90

 towards

0.89

 toward

0.87

not

0.85

NOT

0.84

 everywhere

0.82

 accordingly

0.80

Activations Density 0.135%

words related to urging or advising actions or behaviors

references to groups of people being encouraged or advised to take specific actions

No Comments

No Known Activations