INDEX

Explanations

personalities and titles, particularly those related to government, organizations, or professions

oai_token-act-pair · gpt-3.5-turbo

instances of individuals' names or titles followed by their associated actions or attributions

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

MJ

-0.65

 trivia

-0.64

 ponies

-0.63

 plates

-0.63

 parity

-0.62

 futures

-0.60

 qualifying

-0.58

pace

-0.58

BOX

-0.58

 oppress

-0.58

POSITIVE LOGITS

Jr

0.92

ogun

0.91

hower

0.87

ÄŁ

0.83

jri

0.82

ablishment

0.82

 presided

0.80

 oversaw

0.78

etheless

0.78

jriwal

0.77

Activations Density 0.374%

personalities and titles, particularly those related to government, organizations, or professions

instances of individuals' names or titles followed by their associated actions or attributions

No Comments

No Known Activations