INDEX

Explanations

references to polar or polarized concepts and phenomena

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

bler

-0.82

rahim

-0.79

esis

-0.73

dump

-0.71

hig

-0.66

lished

-0.66

aum

-0.65

CHA

-0.65

ECH

-0.65

RF

-0.64

POSITIVE LOGITS

 vortex

0.97

ities

0.95

 oppos

0.89

oid

0.88

 opposite

0.81

 bear

0.79

izing

0.78

imeter

0.76

igr

0.75

itary

0.75

Activations Density 0.096%

references to polar or polarized concepts and phenomena

No Comments

No Known Activations