INDEX

Explanations

words related to breaking news or events

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "broke."

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

umatic

-0.65

erva

-0.64

accompanied

-0.63

imental

-0.60

enza

-0.60

reau

-0.60

affer

-0.60

iveness

-0.60

iva

-0.58

oft

-0.58

POSITIVE LOGITS

neck

1.10

red

0.88

bre

0.83

 broke

0.79

fast

0.78

breaks

0.74

LEASE

0.73

bones

0.71

bow

0.70

dig

0.69

Activations Density 0.021%

words related to breaking news or events

occurrences of the word "broke."

No Comments

No Known Activations