INDEX

Explanations

words related to barriers or obstacles

oai_token-act-pair · gpt-3.5-turbo

references to gates and gatekeepers, implying control or access points

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Hots

-0.85

ensional

-0.73

yles

-0.70

arus

-0.69

issance

-0.68

ortium

-0.67

 Norm

-0.66

ity

-0.65

urrent

-0.62

itarian

-0.61

POSITIVE LOGITS

keepers

1.38

keeper

1.37

ways

1.21

keeping

1.07

fold

1.06

posts

0.98

way

0.96

stones

0.92

hole

0.90

house

0.85

Activations Density 0.032%

words related to barriers or obstacles

references to gates and gatekeepers, implying control or access points

No Comments

No Known Activations