INDEX

Explanations

words related to information organization and categorization

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a variety of conditions or statuses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

obook

-0.78

itamin

-0.73

iggurat

-0.69

ocrat

-0.69

osate

-0.68

Econom

-0.68

actionDate

-0.68

cellence

-0.67

ractor

-0.67

ukemia

-0.67

POSITIVE LOGITS

 instances

1.44

 examples

1.43

 scenarios

1.42

 versions

1.37

 responses

1.30

 occasions

1.30

 batches

1.30

 setups

1.30

 situations

1.30

 occurrences

1.29

Activations Density 0.711%

words related to information organization and categorization

phrases indicating a variety of conditions or statuses

No Comments

No Known Activations