INDEX

Explanations

terms related to abstract concepts or ideas

oai_token-act-pair · gpt-3.5-turbo

references to abstract concepts or forms in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

odder

-0.73

wreck

-0.71

cture

-0.71

UNCH

-0.69

owship

-0.68

tackle

-0.68

rican

-0.68

rimp

-0.66

ICAN

-0.66

CVE

-0.66

POSITIVE LOGITS

ions

1.33

edly

0.96

urally

0.93

 syntax

0.90

 algebra

0.89

 Expression

0.82

 abstract

0.80

ural

0.80

ed

0.78

ured

0.77

Activations Density 0.038%

terms related to abstract concepts or ideas

references to abstract concepts or forms in various contexts

No Comments

No Known Activations