INDEX

Explanations

years or numerical indicators related to time

oai_token-act-pair · gpt-3.5-turbo

references to specific years

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

 showc

-0.71

ktop

-0.68

 cumbers

-0.65

NetMessage

-0.64

 horr

-0.62

 contag

-0.62

 choke

-0.62

ecast

-0.61

 bluff

-0.61

 footh

-0.60

POSITIVE LOGITS

long

1.16

nings

1.12

book

1.12

ning

1.10

lings

1.02

books

0.95

lies

0.91

 olds

0.90

ns

0.89

Ago

0.86

Activations Density 0.109%

years or numerical indicators related to time

references to specific years

No Comments

No Known Activations