INDEX

Explanations

phrases related to the years 2008 and 2009

oai_token-act-pair · gpt-3.5-turbo

references to specific years, particularly 2008 and 2009

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ez

-0.64

ube

-0.63

 ancest

-0.62

lihood

-0.60

ļ

-0.59

phy

-0.58

ophen

-0.58

estine

-0.57

abulary

-0.57

 Templar

-0.55

POSITIVE LOGITS

-'

1.05

å¹

0.94

 onwards

0.80

shire

0.68

clinton

0.67

âĢ²

0.66

âĪĴ

0.66

iversary

0.65

0.64

Activations Density 0.108%

phrases related to the years 2008 and 2009

references to specific years, particularly 2008 and 2009

No Comments

No Known Activations