INDEX

Explanations

instances where something can be improved or done better

oai_token-act-pair · gpt-3.5-turbo

phrases indicating improvement or positive performance

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

Tru

-0.75

jected

-0.75

 Personality

-0.69

 Mastery

-0.69

went

-0.67

ipel

-0.66

ixed

-0.64

 Cutter

-0.63

shaw

-0.61

ãĤ¦ãĤ¹

-0.61

POSITIVE LOGITS

 grunt

0.78

job

0.75

 injustice

0.74

ingen

0.71

benefit

0.68

 offline

0.68

 homework

0.67

 deed

0.67

eret

0.64

 deserve

0.64

Activations Density 0.077%

instances where something can be improved or done better

phrases indicating improvement or positive performance

No Comments

No Known Activations