INDEX

Explanations

phrases related to things being "undisturbed"

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "und," often indicating themes of negation or lack

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

anwhile

-0.88

azines

-0.84

 Flavoring

-0.82

gdala

-0.82

ebus

-0.81

å§«

-0.81

zzo

-0.80

ppo

-0.80

ramid

-0.80

azine

-0.79

POSITIVE LOGITS

oubt

0.94

iscovered

0.93

aunted

0.91

etermined

0.86

oub

0.85

ated

0.82

ried

0.81

rained

0.80

irect

0.80

rawn

0.80

Activations Density 0.008%

phrases related to things being "undisturbed"

instances of the prefix "und," often indicating themes of negation or lack

No Comments

No Known Activations