INDEX

Explanations

names, specifically the name "Najib Razak"

oai_token-act-pair · gpt-3.5-turbo

proper nouns, specifically names often related to a particular context or narrative

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

sonian

-0.90

gling

-0.79

rome

-0.78

ãĥ¼ãĥĨãĤ£

-0.75

glers

-0.72

 Dull

-0.70

 Circle

-0.65

gest

-0.64

stone

-0.64

 Pioneer

-0.63

POSITIVE LOGITS

ibaba

0.95

eed

0.86

idav

0.84

irmed

0.83

orthy

0.82

velength

0.81

irms

0.81

poons

0.79

iband

0.77

ee

0.76

Activations Density 0.047%

names, specifically the name "Najib Razak"

proper nouns, specifically names often related to a particular context or narrative

No Comments

No Known Activations