INDEX

Explanations

instances where written words convey important information or instructions

oai_token-act-pair · gpt-3.5-turbo

instances of statistical data or references to agreements or disagreements among groups

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Secrets

-0.41

Alert

-0.39

 earthqu

-0.39

 Stupid

-0.38

avanaugh

-0.37

 contrace

-0.37

 idiots

-0.37

 WARN

-0.37

srf

-0.36

CPR

-0.36

POSITIVE LOGITS

 indeed

0.62

 nonetheless

0.59

 quite

0.56

also

0.56

 also

0.56

 albeit

0.54

 furthermore

0.54

 perhaps

0.54

 certainly

0.54

 nevertheless

0.54

Activations Density 1.974%

instances where written words convey important information or instructions

instances of statistical data or references to agreements or disagreements among groups

No Comments

No Known Activations