INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 bind

-0.65

 Shield

-0.61

google

-0.59

 Choose

-0.59

 RELEASE

-0.59

 veto

-0.58

reply

-0.57

 headers

-0.57

<-

-0.57

 transcripts

-0.56

POSITIVE LOGITS

 extreme

1.45

extreme

1.12

 extremes

0.88

nce

0.79

âĶĢâĶĢâĶĢâĶĢ

0.74

 extrem

0.71

ifice

0.70

Site

0.70

perature

0.69

 embr

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.