INDEX

Explanations

mentions of actions or events being attributed to certain people or sources

oai_token-act-pair · gpt-3.5-turbo

phrases indicating attribution or sourcing of information

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

ennial

-0.80

iste

-0.75

NetMessage

-0.72

ãĥĺ

-0.71

bowl

-0.69

nder

-0.69

eless

-0.69

sis

-0.68

Generic

-0.67

Plugin

-0.67

POSITIVE LOGITS

 sources

1.22

 excerpts

1.19

 transcripts

1.06

 documents

1.05

 interviews

1.00

 reports

0.99

 emails

0.95

 filings

0.95

 witnesses

0.94

 insiders

0.94

Activations Density 0.104%

mentions of actions or events being attributed to certain people or sources

phrases indicating attribution or sourcing of information

No Comments

No Known Activations