INDEX

Explanations

instances of the word "mistake" or its variations

oai_token-act-pair · gpt-3.5-turbo

references to errors or mistakes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

electric

-0.83

vez

-0.77

ighth

-0.75

rollers

-0.73

amen

-0.72

well

-0.71

estate

-0.70

zona

-0.70

 helicop

-0.68

irie

-0.66

POSITIVE LOGITS

 mistakes

0.88

 mistake

0.79

 unfocusedRange

0.71

ously

0.69

 Woods

0.68

 Endless

0.68

 Andreas

0.67

steps

0.66

 errors

0.66

%%%%

0.66

Activations Density 0.022%

instances of the word "mistake" or its variations

references to errors or mistakes

No Comments

No Known Activations