INDEX

Explanations

years mentioned in texts

oai_token-act-pair · gpt-3.5-turbo

references to the year 2014

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ishing

-0.89

isher

-0.81

ype

-0.77

abilia

-0.74

ishes

-0.74

warts

-0.73

ishment

-0.72

bip

-0.68

ishers

-0.65

 ship

-0.65

POSITIVE LOGITS

å¹

1.04

assetsadobe

0.82

 onwards

0.82

-'

0.72

0.69

gress

0.68

âĹ¼

0.68

gaard

0.68

âĢ²

0.66

 Sochi

0.65

Activations Density 0.036%

years mentioned in texts

references to the year 2014

No Comments

No Known Activations