INDEX

Explanations

the word "flat" with different contexts like physical attributes, conditions, and descriptions

oai_token-act-pair · gpt-3.5-turbo

references to the concept of flatness

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

EMENT

-0.93

IVERS

-0.92

GBT

-0.85

CRIP

-0.80

emonic

-0.79

 appropri

-0.77

externalActionCode

-0.75

 ANGEL

-0.73

Ago

-0.72

Æ

-0.72

POSITIVE LOGITS

tered

1.17

bed

1.16

lining

1.05

bread

1.04

lined

1.02

ulent

0.99

tering

0.97

ulence

0.97

ric

0.96

stairs

0.93

Activations Density 0.019%

the word "flat" with different contexts like physical attributes, conditions, and descriptions

references to the concept of flatness

No Comments

No Known Activations