INDEX

Explanations

numerical counts or totals

oai_token-act-pair · gpt-3.5-turbo

terms related to counting or recording numbers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

aimon

-0.65

por

-0.63

ASE

-0.62

 Pillar

-0.59

 Galile

-0.59

 Khan

-0.58

Study

-0.57

 friendly

-0.57

Kir

-0.57

 Spit

-0.57

POSITIVE LOGITS

 tally

1.38

enance

1.13

 tallied

0.97

 totals

0.87

eering

0.77

 ledger

0.75

eers

0.72

iom

0.72

ocument

0.72

iculty

0.72

Activations Density 0.014%

numerical counts or totals

terms related to counting or recording numbers

No Comments

No Known Activations